DDRi_rADAGE_Project_Model_implementation.ipynb

{
  "nbformat": 4,
  "nbformat_minor": 0,
  "metadata": {
    "colab": {
      "name": "DDRi rADAGE Project Model implementation.ipynb",
      "provenance": [],
      "collapsed_sections": [],
      "include_colab_link": true
    },
    "kernelspec": {
      "name": "python3",
      "display_name": "Python 3"
    }
  },
  "cells": [
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "view-in-github",
        "colab_type": "text"
      },
      "source": [
        "<a href=\"https://colab.research.google.com/github/qczhao92/ADAGE-on-L1000/blob/main/DDRi_rADAGE_Project_Model_implementation.ipynb\" target=\"_parent\"><img src=\"https://colab.research.google.com/assets/colab-badge.svg\" alt=\"Open In Colab\"/></a>"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "NNuQFbQBGsDv"
      },
      "source": [
        "Discussion notebook: https://udemontreal-my.sharepoint.com/:o:/g/personal/qingchuan_zhao_umontreal_ca/EjBKZT4WPJ5Jkemsdlrw1LIBpz42h8cXQ-Jojda_NAv7xA?e=IdaRn6"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "SvFfTJ-Qvf-x",
        "outputId": "0b2289c2-471e-4210-a36b-d563b2d2be35"
      },
      "source": [
        "colab = True\n",
        "# In case in Colab activate this cell by passing the corect path to the images\n",
        "if colab==True: \n",
        "    from google.colab import drive\n",
        "    drive.mount('/content/drive')\n",
        "    work_dir = 'drive/MyDrive/AIforGenomics2021/ADAGE'\n",
        "else:\n",
        "    work_dir = ''"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Drive already mounted at /content/drive; to attempt to forcibly remount, call drive.mount(\"/content/drive\", force_remount=True).\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kFXztVGrwgqC"
      },
      "source": [
        "import sys, os\n",
        "import time\n",
        "import numpy as np\n",
        "import pandas as pd\n",
        "import torch\n",
        "import torch.nn as nn\n",
        "import torch.optim as optim\n",
        "import torch.autograd as autograd\n",
        "from torch.autograd import Variable\n",
        "from torch.utils.data import Dataset, DataLoader\n",
        "\n",
        "import matplotlib\n",
        "matplotlib.use('Agg')\n",
        "import matplotlib.pyplot as plt"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "nsKi3Q3viZ1A"
      },
      "source": [
        "# set the path of files\n",
        "l5data = \"/content/drive/MyDrive/AIforGenomics2021/ADAGE/GSE70138_Broad_LINCS_Level5_COMPZ_n118050x12328_2017-03-06.gctx\"\n",
        "gene_info_txt = \"/content/drive/MyDrive/AIforGenomics2021/ADAGE/GSE70138_Broad_LINCS_gene_info_2017-03-06.txt\"\n",
        "pert_info_txt = \"/content/drive/MyDrive/AIforGenomics2021/ADAGE/GSE70138_Broad_LINCS_pert_info.txt\"\n",
        "sig_info_txt = \"/content/drive/MyDrive/AIforGenomics2021/ADAGE/GSE70138_Broad_LINCS_sig_info_2017-03-06.txt\"\n",
        "cell_info_txt = \"/content/drive/MyDrive/AIforGenomics2021/ADAGE/GSE70138_Broad_LINCS_cell_info_2017-04-28.txt\""
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Sfx7EK9de3V6"
      },
      "source": [
        "# 0. Download L1000 data\n",
        "download the 2 phase level 5 data to the working directory\n",
        "https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE70138"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Yx4YzYbdiMYx"
      },
      "source": [
        "from tqdm import *\n",
        "import os\n",
        "\n",
        "if not os.path.exists(work_dir):\n",
        "    os.system(f'mkdir -p {work_dir}')\n",
        "\n",
        "if not os.path.isfile(gene_info_txt):\n",
        "    gene_info_link ='ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE70nnn/GSE70138/suppl/GSE70138_Broad_LINCS_gene_info_2017-03-06.txt.gz'\n",
        "    os.system(f'wget -O {gene_info_txt}.gz {gene_info_link}')\n",
        "    os.system(f'gunzip {gene_info_txt}.gz')\n",
        "\n",
        "if not os.path.isfile(pert_info_txt):\n",
        "    pert_info_link ='ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE70nnn/GSE70138/suppl/GSE70138_Broad_LINCS_pert_info.txt.gz'\n",
        "    os.system(f'wget -O {pert_info_txt}.gz {pert_info_link}')\n",
        "    os.system(f'gunzip {pert_info_txt}.gz')\n",
        "\n",
        "if not os.path.isfile(sig_info_txt):\n",
        "    sig_info_link ='ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE70nnn/GSE70138/suppl/GSE70138_Broad_LINCS_sig_info_2017-03-06.txt.gz'\n",
        "    os.system(f'wget -O {sig_info_txt}.gz {sig_info_link}')\n",
        "    os.system(f'gunzip {sig_info_txt}.gz')\n",
        "\n",
        "if not os.path.isfile(cell_info_txt):\n",
        "    cell_info_link ='ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE70nnn/GSE70138/suppl/GSE70138_Broad_LINCS_cell_info_2017-04-28.txt.gz'\n",
        "    os.system(f'wget -O {cell_info_txt}.gz {cell_info_link}')\n",
        "    os.system(f'gunzip {cell_info_txt}.gz')\n",
        "\n",
        "if not os.path.isfile(l5data):\n",
        "    if not os.path.isfile(f'{l5data}.gz'):\n",
        "        print('downloading L1000 level5 data... could take a while')\n",
        "        l5data_link ='ftp://ftp.ncbi.nlm.nih.gov/geo/series/GSE70nnn/GSE70138/suppl/GSE70138_Broad_LINCS_Level5_COMPZ_n118050x12328_2017-03-06.gctx.gz'\n",
        "        os.system(f'wget -O {l5data} {l5data_link}')\n",
        "    else:\n",
        "        print('Zipped L1000 level5 data exits!', f'starting unzip {l5data}.gz')\n",
        "    os.system(f'gunzip {l5data}.gz')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "YxqzLriW4IN7",
        "outputId": "39ba705d-c9ae-4360-8807-fadf5c748f56"
      },
      "source": [
        "!ls /content/drive/MyDrive/AIforGenomics2021/ADAGE -all"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "total 6591142\n",
            "-rw------- 1 root root  540934917 Dec 20  2018  9606.protein.links.v11.0.txt\n",
            "-rw------- 1 root root  276119388 Mar 11 22:15  A549_full_gene_expression.csv\n",
            "-rw------- 1 root root   21735820 Mar  8 19:16  A549_landmark_expression.csv\n",
            "-rw------- 1 root root   33272309 Apr  8 18:02  DDRi_landmark_expression.csv\n",
            "-rw------- 1 root root      11850 Mar  1 18:07  GSE70138_Broad_LINCS_cell_info_2017-04-28.txt\n",
            "-rw------- 1 root root     621887 Mar  1 18:07  GSE70138_Broad_LINCS_gene_info_2017-03-06.txt\n",
            "-rw------- 1 root root 5824120558 Mar  1 18:07  GSE70138_Broad_LINCS_Level5_COMPZ_n118050x12328_2017-03-06.gctx\n",
            "-rw------- 1 root root     219965 Mar  1 18:07  GSE70138_Broad_LINCS_pert_info.txt\n",
            "-rw------- 1 root root   18260895 Mar  1 18:07  GSE70138_Broad_LINCS_sig_info_2017-03-06.txt\n",
            "-rw------- 1 root root      26513 Apr  8 19:46  landmark_gene_STRING_ID.txt\n",
            "-rw------- 1 root root   31139750 Mar 22 18:38  PARPi_landmark_expression.csv\n",
            "-rw------- 1 root root     392615 Mar 11 19:04  string_normal_image.png\n",
            "drwx------ 2 root root       4096 Mar  6 21:02  trained_model\n",
            "-rw------- 1 root root    2466424 Mar 11 22:21 'weight matrices after training'\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "QpQqeTSGgLFf"
      },
      "source": [
        "## take a look of the information files"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "id": "ypRszBXftZr1",
        "outputId": "ffce0f0f-311f-408e-d4cf-c03cfae652f6"
      },
      "source": [
        "\n",
        "\n",
        "gene_info = pd.read_csv(gene_info_txt, sep = '\\t')\n",
        "print(gene_info.shape)\n",
        "gene_info.head()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(12328, 5)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>pr_gene_id</th>\n",
              "      <th>pr_gene_symbol</th>\n",
              "      <th>pr_gene_title</th>\n",
              "      <th>pr_is_lm</th>\n",
              "      <th>pr_is_bing</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>780</td>\n",
              "      <td>DDR1</td>\n",
              "      <td>discoidin domain receptor tyrosine kinase 1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>7849</td>\n",
              "      <td>PAX8</td>\n",
              "      <td>paired box 8</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>2978</td>\n",
              "      <td>GUCA1A</td>\n",
              "      <td>guanylate cyclase activator 1A</td>\n",
              "      <td>0</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>2049</td>\n",
              "      <td>EPHB3</td>\n",
              "      <td>EPH receptor B3</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>2101</td>\n",
              "      <td>ESRRA</td>\n",
              "      <td>estrogen related receptor alpha</td>\n",
              "      <td>0</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   pr_gene_id pr_gene_symbol  ... pr_is_lm  pr_is_bing\n",
              "0         780           DDR1  ...        1           1\n",
              "1        7849           PAX8  ...        1           1\n",
              "2        2978         GUCA1A  ...        0           0\n",
              "3        2049          EPHB3  ...        0           1\n",
              "4        2101          ESRRA  ...        0           1\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "t3JgACdPgIFC",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "f23ae835-d13a-46fa-80b8-231fba45ff39"
      },
      "source": [
        "pert_info = pd.read_csv(pert_info_txt, sep = '\\t')\n",
        "print(pert_info.shape)\n",
        "pert_info.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(2170, 5)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>pert_id</th>\n",
              "      <th>canonical_smiles</th>\n",
              "      <th>inchi_key</th>\n",
              "      <th>pert_iname</th>\n",
              "      <th>pert_type</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>BRD-K70792160</td>\n",
              "      <td>CCN(CC)CCCCN1c2ccccc2Oc2ccc(Cl)cc12</td>\n",
              "      <td>GYBXAGDWMCJZJK-UHFFFAOYSA-N</td>\n",
              "      <td>10-DEBC</td>\n",
              "      <td>trt_cp</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>BRD-K68552125</td>\n",
              "      <td>CCCCCCCCCCCCCC(=O)O[C@@H]1[C@@H](C)[C@]2(O)[C@...</td>\n",
              "      <td>PHEDXBVPIONUQT-RGYGYFBISA-N</td>\n",
              "      <td>phorbol-myristate-acetate</td>\n",
              "      <td>trt_cp</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>BRD-K92301463</td>\n",
              "      <td>CCCCC(C)(C)[C@H](O)\\C=C\\[C@H]1[C@H](O)CC(=O)[C...</td>\n",
              "      <td>QAOBBBBDJSWHMU-WMBBNPMCSA-N</td>\n",
              "      <td>16,16-dimethylprostaglandin-e2</td>\n",
              "      <td>trt_cp</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>BRD-A29731977</td>\n",
              "      <td>CCCCCC(=O)O[C@@]1(CCC2C3CCC4=CC(=O)CC[C@]4(C)C...</td>\n",
              "      <td>DOMWKUIIPQCAJU-JKPPDDDBSA-N</td>\n",
              "      <td>17-hydroxyprogesterone-caproate</td>\n",
              "      <td>trt_cp</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>BRD-K07954936</td>\n",
              "      <td>OC(=O)CCCC[C@@H]1SC[C@@H]2NC(=N)N[C@H]12</td>\n",
              "      <td>WWVANQJRLPIHNS-ZKWXMUAHSA-N</td>\n",
              "      <td>2-iminobiotin</td>\n",
              "      <td>trt_cp</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "         pert_id  ... pert_type\n",
              "0  BRD-K70792160  ...    trt_cp\n",
              "1  BRD-K68552125  ...    trt_cp\n",
              "2  BRD-K92301463  ...    trt_cp\n",
              "3  BRD-A29731977  ...    trt_cp\n",
              "4  BRD-K07954936  ...    trt_cp\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 6
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "id": "3VN3UTKN51fI",
        "outputId": "fc88ca6a-09d1-4276-8211-02ecd557412c"
      },
      "source": [
        "sig_info = pd.read_csv(sig_info_txt, sep = '\\t')\n",
        "print(sig_info.shape)\n",
        "sig_info.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(118050, 8)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>sig_id</th>\n",
              "      <th>pert_id</th>\n",
              "      <th>pert_iname</th>\n",
              "      <th>pert_type</th>\n",
              "      <th>cell_id</th>\n",
              "      <th>pert_idose</th>\n",
              "      <th>pert_itime</th>\n",
              "      <th>distil_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>LJP005_A375_24H:A03</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A03|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>LJP005_A375_24H:A04</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A04|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>LJP005_A375_24H:A05</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A05|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>LJP005_A375_24H:A06</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A06|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>LJP005_A375_24H:A07</td>\n",
              "      <td>BRD-K76908866</td>\n",
              "      <td>CP-724714</td>\n",
              "      <td>trt_cp</td>\n",
              "      <td>A375</td>\n",
              "      <td>10.0 um</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A07|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                sig_id  ...                                          distil_id\n",
              "0  LJP005_A375_24H:A03  ...  LJP005_A375_24H_X1_B19:A03|LJP005_A375_24H_X2_...\n",
              "1  LJP005_A375_24H:A04  ...  LJP005_A375_24H_X1_B19:A04|LJP005_A375_24H_X2_...\n",
              "2  LJP005_A375_24H:A05  ...  LJP005_A375_24H_X1_B19:A05|LJP005_A375_24H_X2_...\n",
              "3  LJP005_A375_24H:A06  ...  LJP005_A375_24H_X1_B19:A06|LJP005_A375_24H_X2_...\n",
              "4  LJP005_A375_24H:A07  ...  LJP005_A375_24H_X1_B19:A07|LJP005_A375_24H_X2_...\n",
              "\n",
              "[5 rows x 8 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 8
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "uLLdwnILEVsK",
        "outputId": "29e8df49-9ef7-46d8-b4f2-14849f9867f4"
      },
      "source": [
        "\n",
        "sig_info[sig_info['pert_iname']=='e3330'].shape\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(0, 8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 22
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "KNc1pXcq3-G9",
        "outputId": "8f0940c5-86ba-4446-d838-f4d04a4a3eb9"
      },
      "source": [
        "#pick up DNA damage repair pathway inhibitor treated gene signatures\n",
        "ola=sig_info[sig_info['pert_iname']=='olaparib'].astype(str)\n",
        "ruca=sig_info[sig_info['pert_iname']=='rucaparib'].astype(str)\n",
        "tala=sig_info[sig_info['pert_iname']=='talazoparib'].astype(str)\n",
        "veli=sig_info[sig_info['pert_iname']=='veliparib'].astype(str)\n",
        "ini=sig_info[sig_info['pert_iname']=='iniparib'].astype(str)\n",
        "lome=sig_info[sig_info['pert_iname']=='lomeguatrib'].astype(str)\n",
        "mito=sig_info[sig_info['pert_iname']=='mitoxantrone'].astype(str)\n",
        "AZD7762=sig_info[sig_info['pert_iname']=='AZD-7762'].astype(str)\n",
        "camp=sig_info[sig_info['pert_iname']=='camptothecin'].astype(str)\n",
        "iri=sig_info[sig_info['pert_iname']=='irinotecan'].astype(str)\n",
        "\n",
        "frames1=(ola, ruca, tala,veli, ini, mito, AZD7762, camp, iri, lome)\n",
        "#PARPi=pd.concat(frames1)\n",
        "DDRi=pd.concat(frames1)\n",
        "DDRi.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1630, 8)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 14
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "id": "ZMLulfNO_ub6",
        "outputId": "5e49d0af-7d84-472f-e792-ab764e461d00"
      },
      "source": [
        "#pick up DMSO samples from the cell lines with PARPi treatment\n",
        "DMSO=sig_info[sig_info['pert_iname']=='DMSO'].astype(str)\n",
        "cell_lines=ola.cell_id.unique()\n",
        "#DMSO_PARPi=DMSO[DMSO['cell_id'].isin(cell_lines)]\n",
        "DMSO_DDRi=DMSO[DMSO['cell_id'].isin(cell_lines)]\n",
        "DMSO_DDRi.head()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>sig_id</th>\n",
              "      <th>pert_id</th>\n",
              "      <th>pert_iname</th>\n",
              "      <th>pert_type</th>\n",
              "      <th>cell_id</th>\n",
              "      <th>pert_idose</th>\n",
              "      <th>pert_itime</th>\n",
              "      <th>distil_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>LJP005_A375_24H:A03</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A03|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>LJP005_A375_24H:A04</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A04|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>LJP005_A375_24H:A05</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A05|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>LJP005_A375_24H:A06</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:A06|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>21</th>\n",
              "      <td>LJP005_A375_24H:B03</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>A375</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP005_A375_24H_X1_B19:B03|LJP005_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                 sig_id  ...                                          distil_id\n",
              "0   LJP005_A375_24H:A03  ...  LJP005_A375_24H_X1_B19:A03|LJP005_A375_24H_X2_...\n",
              "1   LJP005_A375_24H:A04  ...  LJP005_A375_24H_X1_B19:A04|LJP005_A375_24H_X2_...\n",
              "2   LJP005_A375_24H:A05  ...  LJP005_A375_24H_X1_B19:A05|LJP005_A375_24H_X2_...\n",
              "3   LJP005_A375_24H:A06  ...  LJP005_A375_24H_X1_B19:A06|LJP005_A375_24H_X2_...\n",
              "21  LJP005_A375_24H:B03  ...  LJP005_A375_24H_X1_B19:B03|LJP005_A375_24H_X2_...\n",
              "\n",
              "[5 rows x 8 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "id": "_ZJovS6r7rxD",
        "outputId": "13df0c2d-b10f-49e8-bdb5-7121a9835448"
      },
      "source": [
        "#randomly choose about 1500 DMSO samples\n",
        "#DMSO_PARPi_R500=DMSO_PARPi.sample(n=500)\n",
        "#DMSO_PARPi_R500.head()\n",
        "DMSO_DDRi_R1500=DMSO_DDRi.sample(n=1500)\n",
        "DMSO_DDRi_R1500.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>sig_id</th>\n",
              "      <th>pert_id</th>\n",
              "      <th>pert_iname</th>\n",
              "      <th>pert_type</th>\n",
              "      <th>cell_id</th>\n",
              "      <th>pert_idose</th>\n",
              "      <th>pert_itime</th>\n",
              "      <th>distil_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>73679</th>\n",
              "      <td>REP.A013_YAPC_24H:A05</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>YAPC</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>REP.A013_YAPC_24H_X1_B22:A05|REP.A013_YAPC_24H...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>59784</th>\n",
              "      <td>REP.A008_MCF7_24H:F07</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>MCF7</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>REP.A008_MCF7_24H_X1_B24:F07|REP.A008_MCF7_24H...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>107921</th>\n",
              "      <td>REP.A026_PC3_24H:F11</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>PC3</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>REP.A026_PC3_24H_X1_B25:F11|REP.A026_PC3_24H_X...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>88674</th>\n",
              "      <td>REP.A019_HT29_24H:J18</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>HT29</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>REP.A019_HT29_24H_X1_B23:J18|REP.A019_HT29_24H...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>66868</th>\n",
              "      <td>REP.A011_HELA_24H:A04</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>DMSO</td>\n",
              "      <td>ctl_vehicle</td>\n",
              "      <td>HELA</td>\n",
              "      <td>-666</td>\n",
              "      <td>24 h</td>\n",
              "      <td>REP.A011_HELA_24H_X1_B24:A04|REP.A011_HELA_24H...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                       sig_id  ...                                          distil_id\n",
              "73679   REP.A013_YAPC_24H:A05  ...  REP.A013_YAPC_24H_X1_B22:A05|REP.A013_YAPC_24H...\n",
              "59784   REP.A008_MCF7_24H:F07  ...  REP.A008_MCF7_24H_X1_B24:F07|REP.A008_MCF7_24H...\n",
              "107921   REP.A026_PC3_24H:F11  ...  REP.A026_PC3_24H_X1_B25:F11|REP.A026_PC3_24H_X...\n",
              "88674   REP.A019_HT29_24H:J18  ...  REP.A019_HT29_24H_X1_B23:J18|REP.A019_HT29_24H...\n",
              "66868   REP.A011_HELA_24H:A04  ...  REP.A011_HELA_24H_X1_B24:A04|REP.A011_HELA_24H...\n",
              "\n",
              "[5 rows x 8 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 24
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "id": "4r5TLc3x8qop",
        "outputId": "0ce05975-cfc8-4c3c-eeba-0de4b5191524"
      },
      "source": [
        "#generate a new df with PARPi treated samples and 500 randome choosen DMSO samples\n",
        "#frames2=(PARPi, DMSO_PARPi_R500)\n",
        "#parpi_df=pd.concat(frames2)\n",
        "#print(parpi_df.shape)\n",
        "frames2=(DDRi, DMSO_DDRi_R1500)\n",
        "ddri_df=pd.concat(frames2)\n",
        "print(ddri_df.shape)\n",
        "ddri_df.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(3130, 8)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>sig_id</th>\n",
              "      <th>pert_id</th>\n",
              "      <th>pert_iname</th>\n",
              "      <th>pert_type</th>\n",
              "      <th>cell_id</th>\n",
              "      <th>pert_idose</th>\n",
              "      <th>pert_itime</th>\n",
              "      <th>distil_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>24186</th>\n",
              "      <td>LJP008_A375_24H:D01</td>\n",
              "      <td>BRD-K02113016</td>\n",
              "      <td>olaparib</td>\n",
              "      <td>trt_cp</td>\n",
              "      <td>A375</td>\n",
              "      <td>10.0 um</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP008_A375_24H_X1_B20:D01|LJP008_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>24187</th>\n",
              "      <td>LJP008_A375_24H:D02</td>\n",
              "      <td>BRD-K02113016</td>\n",
              "      <td>olaparib</td>\n",
              "      <td>trt_cp</td>\n",
              "      <td>A375</td>\n",
              "      <td>3.33 um</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP008_A375_24H_X1_B20:D02|LJP008_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>24188</th>\n",
              "      <td>LJP008_A375_24H:D03</td>\n",
              "      <td>BRD-K02113016</td>\n",
              "      <td>olaparib</td>\n",
              "      <td>trt_cp</td>\n",
              "      <td>A375</td>\n",
              "      <td>1.11 um</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP008_A375_24H_X1_B20:D03|LJP008_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>24189</th>\n",
              "      <td>LJP008_A375_24H:D04</td>\n",
              "      <td>BRD-K02113016</td>\n",
              "      <td>olaparib</td>\n",
              "      <td>trt_cp</td>\n",
              "      <td>A375</td>\n",
              "      <td>0.37 um</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP008_A375_24H_X1_B20:D04|LJP008_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>24190</th>\n",
              "      <td>LJP008_A375_24H:D05</td>\n",
              "      <td>BRD-K02113016</td>\n",
              "      <td>olaparib</td>\n",
              "      <td>trt_cp</td>\n",
              "      <td>A375</td>\n",
              "      <td>0.12 um</td>\n",
              "      <td>24 h</td>\n",
              "      <td>LJP008_A375_24H_X1_B20:D05|LJP008_A375_24H_X2_...</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                    sig_id  ...                                          distil_id\n",
              "24186  LJP008_A375_24H:D01  ...  LJP008_A375_24H_X1_B20:D01|LJP008_A375_24H_X2_...\n",
              "24187  LJP008_A375_24H:D02  ...  LJP008_A375_24H_X1_B20:D02|LJP008_A375_24H_X2_...\n",
              "24188  LJP008_A375_24H:D03  ...  LJP008_A375_24H_X1_B20:D03|LJP008_A375_24H_X2_...\n",
              "24189  LJP008_A375_24H:D04  ...  LJP008_A375_24H_X1_B20:D04|LJP008_A375_24H_X2_...\n",
              "24190  LJP008_A375_24H:D05  ...  LJP008_A375_24H_X1_B20:D05|LJP008_A375_24H_X2_...\n",
              "\n",
              "[5 rows x 8 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 25
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "V9z-FeX3dcCi"
      },
      "source": [
        "## load L1000 level 5 data (gctx)\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "id": "hzfnOjPgpMg1",
        "outputId": "ece45ab7-49ec-44f3-b3af-88fa2f7e1dd8"
      },
      "source": [
        "landmark_gene_list = gene_info[gene_info['pr_is_lm'] == 1].astype(str)\n",
        "print(landmark_gene_list.shape)\n",
        "landmark_gene_list.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(978, 5)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>pr_gene_id</th>\n",
              "      <th>pr_gene_symbol</th>\n",
              "      <th>pr_gene_title</th>\n",
              "      <th>pr_is_lm</th>\n",
              "      <th>pr_is_bing</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>780</td>\n",
              "      <td>DDR1</td>\n",
              "      <td>discoidin domain receptor tyrosine kinase 1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>7849</td>\n",
              "      <td>PAX8</td>\n",
              "      <td>paired box 8</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25</th>\n",
              "      <td>6193</td>\n",
              "      <td>RPS5</td>\n",
              "      <td>ribosomal protein S5</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>43</th>\n",
              "      <td>23</td>\n",
              "      <td>ABCF1</td>\n",
              "      <td>ATP binding cassette subfamily F member 1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>49</th>\n",
              "      <td>9552</td>\n",
              "      <td>SPAG7</td>\n",
              "      <td>sperm associated antigen 7</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   pr_gene_id pr_gene_symbol  ... pr_is_lm pr_is_bing\n",
              "0         780           DDR1  ...        1          1\n",
              "1        7849           PAX8  ...        1          1\n",
              "25       6193           RPS5  ...        1          1\n",
              "43         23          ABCF1  ...        1          1\n",
              "49       9552          SPAG7  ...        1          1\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 45
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "7bx3z__YwPc3",
        "outputId": "8b8c3dc4-7d4a-4454-e35b-2214e29ab08e"
      },
      "source": [
        "!pip install cmapPy"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting cmapPy\n",
            "\u001b[?25l  Downloading https://files.pythonhosted.org/packages/e8/e1/9e90b0dc2051f4a3a2b9553b254889e93a2a3c7a3b925433c6298d99d596/cmapPy-4.0.1-py2.py3-none-any.whl (150kB)\n",
            "\r\u001b[K     |██▏                             | 10kB 14.7MB/s eta 0:00:01\r\u001b[K     |████▍                           | 20kB 20.3MB/s eta 0:00:01\r\u001b[K     |██████▌                         | 30kB 16.0MB/s eta 0:00:01\r\u001b[K     |████████▊                       | 40kB 11.8MB/s eta 0:00:01\r\u001b[K     |███████████                     | 51kB 8.3MB/s eta 0:00:01\r\u001b[K     |█████████████                   | 61kB 7.6MB/s eta 0:00:01\r\u001b[K     |███████████████▎                | 71kB 8.5MB/s eta 0:00:01\r\u001b[K     |█████████████████▌              | 81kB 9.4MB/s eta 0:00:01\r\u001b[K     |███████████████████▋            | 92kB 8.6MB/s eta 0:00:01\r\u001b[K     |█████████████████████▉          | 102kB 8.4MB/s eta 0:00:01\r\u001b[K     |████████████████████████        | 112kB 8.4MB/s eta 0:00:01\r\u001b[K     |██████████████████████████▏     | 122kB 8.4MB/s eta 0:00:01\r\u001b[K     |████████████████████████████▍   | 133kB 8.4MB/s eta 0:00:01\r\u001b[K     |██████████████████████████████▌ | 143kB 8.4MB/s eta 0:00:01\r\u001b[K     |████████████████████████████████| 153kB 8.4MB/s \n",
            "\u001b[?25hRequirement already satisfied: numpy>=1.11.2 in /usr/local/lib/python3.7/dist-packages (from cmapPy) (1.19.5)\n",
            "Requirement already satisfied: h5py>=2.6.0 in /usr/local/lib/python3.7/dist-packages (from cmapPy) (2.10.0)\n",
            "Requirement already satisfied: pandas>=0.18 in /usr/local/lib/python3.7/dist-packages (from cmapPy) (1.1.5)\n",
            "Requirement already satisfied: six in /usr/local/lib/python3.7/dist-packages (from cmapPy) (1.15.0)\n",
            "Requirement already satisfied: requests>=2.13.0 in /usr/local/lib/python3.7/dist-packages (from cmapPy) (2.23.0)\n",
            "Requirement already satisfied: pytz>=2017.2 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.18->cmapPy) (2018.9)\n",
            "Requirement already satisfied: python-dateutil>=2.7.3 in /usr/local/lib/python3.7/dist-packages (from pandas>=0.18->cmapPy) (2.8.1)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests>=2.13.0->cmapPy) (2.10)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests>=2.13.0->cmapPy) (1.24.3)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests>=2.13.0->cmapPy) (2020.12.5)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests>=2.13.0->cmapPy) (3.0.4)\n",
            "Installing collected packages: cmapPy\n",
            "Successfully installed cmapPy-4.0.1\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "MsRyQp_90sy7"
      },
      "source": [
        "## load the full gene expression data\n",
        "from cmapPy.pandasGEXpress.parse import parse\n",
        "#gene_expr_matrix_full = parse(l5data).data_df.T\n",
        "# del gene_expr_matrix_full  ## release RAM"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "yFEHJqP1dYhG"
      },
      "source": [
        "## only load landmark genes\n",
        "from cmapPy.pandasGEXpress.parse import parse\n",
        "\n",
        "## Load the matrix of differential gene expressions as a pandas dataframe\n",
        "gene_expr_matrix = parse(l5data, rid=landmark_gene_list[\"pr_gene_id\"]).data_df.T\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "Qr_itRri_mEi",
        "outputId": "cf9348ac-ed2f-49ae-afc0-3c20151a9151"
      },
      "source": [
        "#extract the PARPi_df landmark gene expression\n",
        "#PARPi_gene_expr = gene_expr_matrix.loc[gene_expr_matrix.index.isin(parpi_df['sig_id'])]\n",
        "#print(PARPi_gene_expr.shape)\n",
        "#save PARPi_gene_expr\n",
        "#PARPi_gene_expr.to_csv(f\"{work_dir}/PARPi_landmark_expression.csv\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(940, 978)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "VR8ZH-LhOHyJ",
        "outputId": "317d72e4-a758-4e9a-f068-ff6960ce6c89"
      },
      "source": [
        "#extract the DDRi_df landmark gene expression\n",
        "DDRi_gene_expr = gene_expr_matrix.loc[gene_expr_matrix.index.isin(ddri_df['sig_id'])]\n",
        "print(DDRi_gene_expr.shape)\n",
        "#save DDRi_gene_expr\n",
        "DDRi_gene_expr.to_csv(f\"{work_dir}/DDRi_landmark_expression.csv\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(3130, 978)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "rK_HN1A832Nc"
      },
      "source": [
        "# 1. ADAGE model\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n",
        "\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "zCfSBTJj3lMR"
      },
      "source": [
        "construct an ADAGE-based model:\n",
        "1. adding random noise to the input data\n",
        "2. training a neural network with hidden nodes that were able to remove added noise to reconstruct the initial data"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 817
        },
        "id": "pwSq1PHu3Wdh",
        "outputId": "6954d926-e04e-4b4f-a47d-1f411d1628b2"
      },
      "source": [
        "from IPython.display import Image\n",
        "Image(work_dir+'/img/ADAGEmodel.png', width=\"800\")"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "image/png": "drive/MyDrive/AIforGenomics2021/ADAGE/img/ADAGEmodel.png",
            "text/plain": [
              "<IPython.core.display.Image object>"
            ]
          },
          "metadata": {
            "tags": [],
            "image/png": {
              "width": "800"
            }
          },
          "execution_count": 11
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "aQFqIYouFGH6"
      },
      "source": [
        "### 1.1 define the model\n",
        "\n",
        "for each sample $x$:\n",
        "- random corrupt a percentage of the genes -> 0\n",
        "- use corrupt sample $x'$ as input of denoising autoencoder\n",
        "\n",
        "for each encoder node, calculate activity with weight matrix W:\n",
        "> $Y = s(Wx'+b)$\n",
        "\n",
        "the reconstructed gene expression by decoder node:\n",
        "> $z =  W'y + b' $\n",
        "\n",
        "use Mean Squared Error as loss function\n",
        "> $L(x,z) = \\frac{1}{n}\\sum_{k=1}^n (x_k  -z_k)^2$\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "5Rv4TUB67Wn8"
      },
      "source": [
        "\n",
        "class Adage_linear_decoder(nn.Module):\n",
        "    \"\"\"Denoising Auto-Encoder: to reconstruct the input from a corrupted gene expression\"\"\"\n",
        "    \n",
        "    def __init__(self, gene_number, node_number = 50, encoder_layers = 1):\n",
        "        super(Adage_linear_decoder, self).__init__()\n",
        "\n",
        "        hidden_layers = []\n",
        "        current_node=gene_number\n",
        "        if encoder_layers > 1:\n",
        "            delta = (gene_number - node_number) / encoder_layers\n",
        "            current_node = gene_number\n",
        "            for i in range(encoder_layers-1):\n",
        "                n = int(current_node - delta)\n",
        "                hidden_layers.append(nn.Linear(current_node, n))\n",
        "                hidden_layers.append(nn.ReLU(inplace=True))\n",
        "                current_node = n\n",
        "\n",
        "        self.encoder = nn.Sequential(\n",
        "            *hidden_layers,\n",
        "            nn.Linear(current_node, node_number),\n",
        "            nn.ReLU(inplace=True)         \n",
        "        )\n",
        "        self.decoder = nn.Linear(node_number, gene_number)\n",
        "        \n",
        "    def forward(self,x):\n",
        "        y = self.encoder(x)\n",
        "        y = self.decoder(y)\n",
        "        return y"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "ZNDIOZo_CFiH"
      },
      "source": [
        "Some functions for training:"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "4f3BiN5vFE5C"
      },
      "source": [
        "# define an random corruption function\n",
        "def random_corruption(train_batch, corruption_rate = 0.05):\n",
        "    \"\"\" Take a pandas dataframe as input. In each sample (as rows), random corrupt a proportion (corruption_rate) of the genes to 0. \"\"\"\n",
        "    nrow, ncol = input_dataframe.shape\n",
        "    corrupt_mask = numpy.random.choice([0, 1], size=ncol, p=[corruption_rate, 1-corruption_rate])\n",
        "\n",
        "    for i in range(1, nrow):\n",
        "        corrupt_array = numpy.random.choice([0, 1], size=ncol, p=[corruption_rate, 1-corruption_rate])\n",
        "        corrupt_mask = np.vstack(corrupt_mask, corrupt_array)\n",
        "    \n",
        "    return input_dataframe.dot(corrupt_mask)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JlprVTfDaysv"
      },
      "source": [
        "# define an random noise introducer function\n",
        "def random_noise(train_batch, random_rate = 0.05, std=0.1):\n",
        "  noise=np.random.normal(0,std,train_batch.shape)\n",
        "  rate=np.random.choice([0, 1], size=train_batch.shape, p=[1-random_rate, random_rate])\n",
        "  new_signal=train_batch+noise*rate\n",
        "  return new_signal\n",
        "    "
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "R1Pq9n3fd2iz"
      },
      "source": [
        "## function to split the training and validation data\n",
        "def train_valid_split(dataframe, train_portion = 0.7):\n",
        "    \"\"\"take a pandas dataframe as input, output two numpy arrays according to train_portion\"\"\"\n",
        "    data = dataframe.to_numpy()\n",
        "    np.random.shuffle(data)\n",
        "    train_data, valid_data = data[:int(data.shape[0]*train_portion),:], data[int(data.shape[0]*train_portion):,:]\n",
        "\n",
        "    print('samples in training data:',train_data.shape[0],'\\nsamples in validation data:',valid_data.shape[0])\n",
        "    return train_data, valid_data\n",
        " \n",
        "## function to prepare input data\n",
        "def dataloader(data, batch_size, reuse = True):\n",
        "    \"\"\" prepare a batch of datasets for ADAGE model input \"\"\"\n",
        "    np.random.shuffle(data)\n",
        "    num_batches = len(data) // batch_size\n",
        "    while True:\n",
        "        for i in range(num_batches):\n",
        "            yield data[i * batch_size : (i + 1) * batch_size]\n",
        "        if not reuse and i == num_batches - 1:\n",
        "            yield None\n",
        "\n",
        "## function to save model\n",
        "def save_models(work_dir, model, epoch_number,model_name):\n",
        "    \"\"\"Function to save trained model\"\"\"\n",
        "    save_dir = os.path.join(work_dir, \"trained_model\")\n",
        "    os.makedirs(save_dir, exist_ok=True)\n",
        "    torch.save(model.state_dict(), '{}/{}_epoch{}'.format(save_dir, model_name, epoch_number))\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "vGud6CysE1_I"
      },
      "source": [
        "### 1.2 train the model"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Ekj7oqEICMqx"
      },
      "source": [
        "load the data for training and validation"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 301
        },
        "id": "XvDVg0iEjTK5",
        "outputId": "c4d0992e-4d8d-4b6d-e969-38eaf5673c11"
      },
      "source": [
        "#PARPi_gene_expr = pd.read_csv(f\"{work_dir}/PARPi_landmark_expression.csv\", index_col=0)\n",
        "\n",
        "#print(f'data range from {PARPi_gene_expr.values.min()} to {PARPi_gene_expr.values.max()}')\n",
        "#PARPi_gene_expr.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data range from -10.000001 to 10.000001\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>780</th>\n",
              "      <th>7849</th>\n",
              "      <th>6193</th>\n",
              "      <th>23</th>\n",
              "      <th>9552</th>\n",
              "      <th>387</th>\n",
              "      <th>10921</th>\n",
              "      <th>10285</th>\n",
              "      <th>533</th>\n",
              "      <th>6194</th>\n",
              "      <th>23326</th>\n",
              "      <th>351</th>\n",
              "      <th>1213</th>\n",
              "      <th>9761</th>\n",
              "      <th>1465</th>\n",
              "      <th>808</th>\n",
              "      <th>5792</th>\n",
              "      <th>3337</th>\n",
              "      <th>7494</th>\n",
              "      <th>2896</th>\n",
              "      <th>3098</th>\n",
              "      <th>11014</th>\n",
              "      <th>6500</th>\n",
              "      <th>823</th>\n",
              "      <th>813</th>\n",
              "      <th>1509</th>\n",
              "      <th>4144</th>\n",
              "      <th>3925</th>\n",
              "      <th>1891</th>\n",
              "      <th>8826</th>\n",
              "      <th>3303</th>\n",
              "      <th>3329</th>\n",
              "      <th>1153</th>\n",
              "      <th>5720</th>\n",
              "      <th>5048</th>\n",
              "      <th>10525</th>\n",
              "      <th>5708</th>\n",
              "      <th>2058</th>\n",
              "      <th>5710</th>\n",
              "      <th>5223</th>\n",
              "      <th>...</th>\n",
              "      <th>11182</th>\n",
              "      <th>63933</th>\n",
              "      <th>54850</th>\n",
              "      <th>29916</th>\n",
              "      <th>55179</th>\n",
              "      <th>29083</th>\n",
              "      <th>7099</th>\n",
              "      <th>79961</th>\n",
              "      <th>55825</th>\n",
              "      <th>51742</th>\n",
              "      <th>10818</th>\n",
              "      <th>81533</th>\n",
              "      <th>665</th>\n",
              "      <th>10776</th>\n",
              "      <th>64422</th>\n",
              "      <th>10813</th>\n",
              "      <th>80349</th>\n",
              "      <th>1978</th>\n",
              "      <th>83743</th>\n",
              "      <th>57406</th>\n",
              "      <th>23410</th>\n",
              "      <th>8996</th>\n",
              "      <th>55620</th>\n",
              "      <th>23597</th>\n",
              "      <th>124583</th>\n",
              "      <th>54915</th>\n",
              "      <th>3157</th>\n",
              "      <th>85377</th>\n",
              "      <th>55793</th>\n",
              "      <th>466</th>\n",
              "      <th>54681</th>\n",
              "      <th>11000</th>\n",
              "      <th>6915</th>\n",
              "      <th>6253</th>\n",
              "      <th>7264</th>\n",
              "      <th>5467</th>\n",
              "      <th>2767</th>\n",
              "      <th>23038</th>\n",
              "      <th>57048</th>\n",
              "      <th>79716</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cid</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:J15</th>\n",
              "      <td>0.547654</td>\n",
              "      <td>0.395950</td>\n",
              "      <td>0.461493</td>\n",
              "      <td>-0.321076</td>\n",
              "      <td>0.171693</td>\n",
              "      <td>0.679893</td>\n",
              "      <td>-0.116500</td>\n",
              "      <td>-0.142482</td>\n",
              "      <td>-0.018586</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>-0.259256</td>\n",
              "      <td>-0.266754</td>\n",
              "      <td>0.013334</td>\n",
              "      <td>0.615622</td>\n",
              "      <td>0.342822</td>\n",
              "      <td>-0.015804</td>\n",
              "      <td>0.379034</td>\n",
              "      <td>-0.082581</td>\n",
              "      <td>-0.490875</td>\n",
              "      <td>0.529229</td>\n",
              "      <td>1.097920</td>\n",
              "      <td>0.083174</td>\n",
              "      <td>0.223464</td>\n",
              "      <td>0.822543</td>\n",
              "      <td>-0.646200</td>\n",
              "      <td>0.105072</td>\n",
              "      <td>0.386202</td>\n",
              "      <td>0.611098</td>\n",
              "      <td>-0.215986</td>\n",
              "      <td>-0.233854</td>\n",
              "      <td>-0.489597</td>\n",
              "      <td>-0.674382</td>\n",
              "      <td>0.746366</td>\n",
              "      <td>0.424303</td>\n",
              "      <td>-0.134211</td>\n",
              "      <td>0.217497</td>\n",
              "      <td>-1.080208</td>\n",
              "      <td>0.647623</td>\n",
              "      <td>0.420543</td>\n",
              "      <td>-4.904697</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.091081</td>\n",
              "      <td>0.325788</td>\n",
              "      <td>-0.106197</td>\n",
              "      <td>0.603988</td>\n",
              "      <td>0.292633</td>\n",
              "      <td>-0.251299</td>\n",
              "      <td>-0.411023</td>\n",
              "      <td>0.837104</td>\n",
              "      <td>-0.876019</td>\n",
              "      <td>0.427095</td>\n",
              "      <td>-0.398287</td>\n",
              "      <td>-0.047654</td>\n",
              "      <td>0.112173</td>\n",
              "      <td>0.414505</td>\n",
              "      <td>-3.028268</td>\n",
              "      <td>0.472928</td>\n",
              "      <td>0.539418</td>\n",
              "      <td>-0.339128</td>\n",
              "      <td>-0.505857</td>\n",
              "      <td>-0.134691</td>\n",
              "      <td>-1.025917</td>\n",
              "      <td>-0.073245</td>\n",
              "      <td>-0.022658</td>\n",
              "      <td>-5.064618</td>\n",
              "      <td>0.100468</td>\n",
              "      <td>0.423215</td>\n",
              "      <td>0.119007</td>\n",
              "      <td>0.955357</td>\n",
              "      <td>-0.564962</td>\n",
              "      <td>-0.446641</td>\n",
              "      <td>-0.124934</td>\n",
              "      <td>0.664942</td>\n",
              "      <td>0.527476</td>\n",
              "      <td>0.352047</td>\n",
              "      <td>-0.052228</td>\n",
              "      <td>-0.784120</td>\n",
              "      <td>-0.967061</td>\n",
              "      <td>-0.393171</td>\n",
              "      <td>0.398492</td>\n",
              "      <td>-0.336368</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:J16</th>\n",
              "      <td>-0.011768</td>\n",
              "      <td>-1.067306</td>\n",
              "      <td>-0.113229</td>\n",
              "      <td>0.402519</td>\n",
              "      <td>-0.256621</td>\n",
              "      <td>-0.536988</td>\n",
              "      <td>0.471351</td>\n",
              "      <td>-0.789072</td>\n",
              "      <td>0.221237</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>0.597305</td>\n",
              "      <td>0.131708</td>\n",
              "      <td>0.697712</td>\n",
              "      <td>0.274119</td>\n",
              "      <td>-0.225368</td>\n",
              "      <td>0.523252</td>\n",
              "      <td>0.356316</td>\n",
              "      <td>-0.231118</td>\n",
              "      <td>-1.101972</td>\n",
              "      <td>1.162249</td>\n",
              "      <td>0.326193</td>\n",
              "      <td>-0.326843</td>\n",
              "      <td>0.248999</td>\n",
              "      <td>0.237429</td>\n",
              "      <td>-1.456901</td>\n",
              "      <td>-0.219513</td>\n",
              "      <td>0.771570</td>\n",
              "      <td>0.128634</td>\n",
              "      <td>0.646212</td>\n",
              "      <td>-0.020452</td>\n",
              "      <td>-3.920552</td>\n",
              "      <td>-0.772525</td>\n",
              "      <td>2.329190</td>\n",
              "      <td>-0.489395</td>\n",
              "      <td>-0.108981</td>\n",
              "      <td>0.183556</td>\n",
              "      <td>0.053364</td>\n",
              "      <td>-0.193271</td>\n",
              "      <td>0.233431</td>\n",
              "      <td>0.556484</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.326120</td>\n",
              "      <td>-0.207436</td>\n",
              "      <td>-0.190311</td>\n",
              "      <td>-0.273263</td>\n",
              "      <td>-0.475273</td>\n",
              "      <td>-0.218992</td>\n",
              "      <td>-0.173596</td>\n",
              "      <td>-0.067220</td>\n",
              "      <td>-0.734452</td>\n",
              "      <td>1.009016</td>\n",
              "      <td>-1.384133</td>\n",
              "      <td>-0.524755</td>\n",
              "      <td>-0.227863</td>\n",
              "      <td>0.474416</td>\n",
              "      <td>-0.320282</td>\n",
              "      <td>-0.850814</td>\n",
              "      <td>1.125106</td>\n",
              "      <td>-0.108594</td>\n",
              "      <td>-0.597689</td>\n",
              "      <td>0.287199</td>\n",
              "      <td>-0.506501</td>\n",
              "      <td>-0.469675</td>\n",
              "      <td>-1.182983</td>\n",
              "      <td>-0.073488</td>\n",
              "      <td>0.429464</td>\n",
              "      <td>1.339783</td>\n",
              "      <td>0.272432</td>\n",
              "      <td>-0.030471</td>\n",
              "      <td>0.859062</td>\n",
              "      <td>-0.638941</td>\n",
              "      <td>0.079704</td>\n",
              "      <td>-0.052292</td>\n",
              "      <td>0.420883</td>\n",
              "      <td>0.174849</td>\n",
              "      <td>0.901950</td>\n",
              "      <td>-0.674315</td>\n",
              "      <td>-0.665947</td>\n",
              "      <td>0.821813</td>\n",
              "      <td>0.212107</td>\n",
              "      <td>-0.332485</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_HA1E_24H:A03</th>\n",
              "      <td>-0.612091</td>\n",
              "      <td>0.855054</td>\n",
              "      <td>-0.557836</td>\n",
              "      <td>-0.036146</td>\n",
              "      <td>-0.533345</td>\n",
              "      <td>0.430169</td>\n",
              "      <td>0.282822</td>\n",
              "      <td>0.654875</td>\n",
              "      <td>-0.239361</td>\n",
              "      <td>0.768740</td>\n",
              "      <td>0.392429</td>\n",
              "      <td>-0.381888</td>\n",
              "      <td>0.759375</td>\n",
              "      <td>-0.328399</td>\n",
              "      <td>0.399033</td>\n",
              "      <td>-0.758860</td>\n",
              "      <td>0.309794</td>\n",
              "      <td>0.802563</td>\n",
              "      <td>0.908873</td>\n",
              "      <td>-0.506241</td>\n",
              "      <td>0.042305</td>\n",
              "      <td>0.278327</td>\n",
              "      <td>-0.636894</td>\n",
              "      <td>-0.522274</td>\n",
              "      <td>0.305060</td>\n",
              "      <td>-0.587657</td>\n",
              "      <td>0.357382</td>\n",
              "      <td>0.270920</td>\n",
              "      <td>0.212523</td>\n",
              "      <td>-0.301421</td>\n",
              "      <td>-0.553881</td>\n",
              "      <td>0.440242</td>\n",
              "      <td>-0.092588</td>\n",
              "      <td>0.224921</td>\n",
              "      <td>0.794614</td>\n",
              "      <td>0.587701</td>\n",
              "      <td>0.790196</td>\n",
              "      <td>0.767115</td>\n",
              "      <td>0.097447</td>\n",
              "      <td>0.852140</td>\n",
              "      <td>...</td>\n",
              "      <td>0.575854</td>\n",
              "      <td>0.416091</td>\n",
              "      <td>0.335926</td>\n",
              "      <td>-0.105096</td>\n",
              "      <td>-1.305800</td>\n",
              "      <td>0.154202</td>\n",
              "      <td>-0.662583</td>\n",
              "      <td>-0.379384</td>\n",
              "      <td>-1.224841</td>\n",
              "      <td>0.785541</td>\n",
              "      <td>-0.255037</td>\n",
              "      <td>-0.050673</td>\n",
              "      <td>-0.406081</td>\n",
              "      <td>0.352848</td>\n",
              "      <td>0.547996</td>\n",
              "      <td>1.790532</td>\n",
              "      <td>0.505057</td>\n",
              "      <td>-0.638764</td>\n",
              "      <td>-0.674412</td>\n",
              "      <td>-0.240601</td>\n",
              "      <td>-0.084365</td>\n",
              "      <td>-0.153685</td>\n",
              "      <td>0.451601</td>\n",
              "      <td>0.005528</td>\n",
              "      <td>0.715036</td>\n",
              "      <td>0.149368</td>\n",
              "      <td>1.115302</td>\n",
              "      <td>0.216178</td>\n",
              "      <td>4.816076</td>\n",
              "      <td>-0.433131</td>\n",
              "      <td>0.153344</td>\n",
              "      <td>1.064444</td>\n",
              "      <td>-0.119862</td>\n",
              "      <td>0.250326</td>\n",
              "      <td>-0.233840</td>\n",
              "      <td>-0.070832</td>\n",
              "      <td>0.092717</td>\n",
              "      <td>-0.193244</td>\n",
              "      <td>-0.297614</td>\n",
              "      <td>1.408907</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_HA1E_24H:F08</th>\n",
              "      <td>-0.295567</td>\n",
              "      <td>0.264896</td>\n",
              "      <td>0.225759</td>\n",
              "      <td>0.416773</td>\n",
              "      <td>0.554485</td>\n",
              "      <td>0.319323</td>\n",
              "      <td>0.632860</td>\n",
              "      <td>0.218394</td>\n",
              "      <td>0.062931</td>\n",
              "      <td>-1.405682</td>\n",
              "      <td>-0.888978</td>\n",
              "      <td>-0.092187</td>\n",
              "      <td>-0.783735</td>\n",
              "      <td>0.982590</td>\n",
              "      <td>0.630364</td>\n",
              "      <td>0.062697</td>\n",
              "      <td>-0.597367</td>\n",
              "      <td>-0.245464</td>\n",
              "      <td>0.436860</td>\n",
              "      <td>0.747253</td>\n",
              "      <td>0.158223</td>\n",
              "      <td>-0.070773</td>\n",
              "      <td>0.817261</td>\n",
              "      <td>0.522098</td>\n",
              "      <td>-0.046936</td>\n",
              "      <td>0.105714</td>\n",
              "      <td>0.296335</td>\n",
              "      <td>0.354037</td>\n",
              "      <td>-0.358646</td>\n",
              "      <td>-0.606546</td>\n",
              "      <td>-0.873109</td>\n",
              "      <td>0.355777</td>\n",
              "      <td>1.321212</td>\n",
              "      <td>0.378128</td>\n",
              "      <td>0.302558</td>\n",
              "      <td>-0.103678</td>\n",
              "      <td>1.104395</td>\n",
              "      <td>0.535756</td>\n",
              "      <td>0.772063</td>\n",
              "      <td>0.209416</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.569899</td>\n",
              "      <td>0.037671</td>\n",
              "      <td>-0.080219</td>\n",
              "      <td>0.677563</td>\n",
              "      <td>0.377883</td>\n",
              "      <td>0.585444</td>\n",
              "      <td>-0.471824</td>\n",
              "      <td>0.461182</td>\n",
              "      <td>0.325555</td>\n",
              "      <td>-0.321512</td>\n",
              "      <td>-0.341110</td>\n",
              "      <td>-1.335096</td>\n",
              "      <td>0.882909</td>\n",
              "      <td>0.597172</td>\n",
              "      <td>0.188407</td>\n",
              "      <td>-0.172315</td>\n",
              "      <td>0.318651</td>\n",
              "      <td>0.045284</td>\n",
              "      <td>0.087886</td>\n",
              "      <td>-0.644419</td>\n",
              "      <td>4.048363</td>\n",
              "      <td>0.274434</td>\n",
              "      <td>-0.137777</td>\n",
              "      <td>0.280535</td>\n",
              "      <td>0.772253</td>\n",
              "      <td>0.349798</td>\n",
              "      <td>-0.569966</td>\n",
              "      <td>-0.644179</td>\n",
              "      <td>2.285106</td>\n",
              "      <td>0.370050</td>\n",
              "      <td>1.895253</td>\n",
              "      <td>-0.882176</td>\n",
              "      <td>-0.299036</td>\n",
              "      <td>-0.871814</td>\n",
              "      <td>0.408248</td>\n",
              "      <td>0.401330</td>\n",
              "      <td>-0.482904</td>\n",
              "      <td>3.989376</td>\n",
              "      <td>0.614656</td>\n",
              "      <td>-0.309702</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_HA1E_24H:J14</th>\n",
              "      <td>-0.844051</td>\n",
              "      <td>0.076854</td>\n",
              "      <td>0.458108</td>\n",
              "      <td>-0.285793</td>\n",
              "      <td>-0.434120</td>\n",
              "      <td>0.084801</td>\n",
              "      <td>0.670759</td>\n",
              "      <td>0.649724</td>\n",
              "      <td>0.291303</td>\n",
              "      <td>0.319341</td>\n",
              "      <td>-1.645912</td>\n",
              "      <td>-0.476368</td>\n",
              "      <td>1.074345</td>\n",
              "      <td>1.229127</td>\n",
              "      <td>0.949046</td>\n",
              "      <td>-0.447085</td>\n",
              "      <td>-0.387231</td>\n",
              "      <td>0.096495</td>\n",
              "      <td>-0.031915</td>\n",
              "      <td>0.389765</td>\n",
              "      <td>0.951611</td>\n",
              "      <td>0.625000</td>\n",
              "      <td>1.182070</td>\n",
              "      <td>1.178186</td>\n",
              "      <td>1.350742</td>\n",
              "      <td>-0.595371</td>\n",
              "      <td>-0.105703</td>\n",
              "      <td>-0.129478</td>\n",
              "      <td>-0.705864</td>\n",
              "      <td>-0.629163</td>\n",
              "      <td>-1.338078</td>\n",
              "      <td>0.705596</td>\n",
              "      <td>0.815922</td>\n",
              "      <td>0.940423</td>\n",
              "      <td>-0.513836</td>\n",
              "      <td>0.025810</td>\n",
              "      <td>-0.146347</td>\n",
              "      <td>0.731371</td>\n",
              "      <td>0.663340</td>\n",
              "      <td>0.919985</td>\n",
              "      <td>...</td>\n",
              "      <td>0.699243</td>\n",
              "      <td>-0.567127</td>\n",
              "      <td>-0.260456</td>\n",
              "      <td>0.568472</td>\n",
              "      <td>0.468126</td>\n",
              "      <td>-0.130796</td>\n",
              "      <td>0.806007</td>\n",
              "      <td>-0.128329</td>\n",
              "      <td>-0.798106</td>\n",
              "      <td>0.040133</td>\n",
              "      <td>-0.289242</td>\n",
              "      <td>-1.008751</td>\n",
              "      <td>0.614915</td>\n",
              "      <td>0.347262</td>\n",
              "      <td>-0.374585</td>\n",
              "      <td>1.429297</td>\n",
              "      <td>0.720131</td>\n",
              "      <td>0.606112</td>\n",
              "      <td>0.816385</td>\n",
              "      <td>-1.055574</td>\n",
              "      <td>0.870154</td>\n",
              "      <td>-1.016624</td>\n",
              "      <td>-0.952067</td>\n",
              "      <td>0.945955</td>\n",
              "      <td>-1.520199</td>\n",
              "      <td>0.631701</td>\n",
              "      <td>0.905047</td>\n",
              "      <td>0.730261</td>\n",
              "      <td>-0.016199</td>\n",
              "      <td>0.554783</td>\n",
              "      <td>1.024556</td>\n",
              "      <td>-0.749667</td>\n",
              "      <td>-0.314004</td>\n",
              "      <td>-1.098928</td>\n",
              "      <td>0.420196</td>\n",
              "      <td>0.577327</td>\n",
              "      <td>0.824373</td>\n",
              "      <td>0.272414</td>\n",
              "      <td>1.301462</td>\n",
              "      <td>-0.430687</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 978 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                            780      7849  ...     57048     79716\n",
              "cid                                        ...                    \n",
              "REP.A001_A375_24H:J15  0.547654  0.395950  ...  0.398492 -0.336368\n",
              "REP.A001_A375_24H:J16 -0.011768 -1.067306  ...  0.212107 -0.332485\n",
              "REP.A001_HA1E_24H:A03 -0.612091  0.855054  ... -0.297614  1.408907\n",
              "REP.A001_HA1E_24H:F08 -0.295567  0.264896  ...  0.614656 -0.309702\n",
              "REP.A001_HA1E_24H:J14 -0.844051  0.076854  ...  1.301462 -0.430687\n",
              "\n",
              "[5 rows x 978 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 85
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 301
        },
        "id": "50z4-6RPPJa1",
        "outputId": "a35c5ab6-99c2-4b96-b7d8-8429f9c97ff4"
      },
      "source": [
        "DDRi_gene_expr = pd.read_csv(f\"{work_dir}/DDRi_landmark_expression.csv\", index_col=0)\n",
        "\n",
        "print(f'data range from {DDRi_gene_expr.values.min()} to {DDRi_gene_expr.values.max()}')\n",
        "DDRi_gene_expr.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "data range from -10.000001 to 10.000001\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>780</th>\n",
              "      <th>7849</th>\n",
              "      <th>6193</th>\n",
              "      <th>23</th>\n",
              "      <th>9552</th>\n",
              "      <th>387</th>\n",
              "      <th>10921</th>\n",
              "      <th>10285</th>\n",
              "      <th>533</th>\n",
              "      <th>6194</th>\n",
              "      <th>23326</th>\n",
              "      <th>351</th>\n",
              "      <th>1213</th>\n",
              "      <th>9761</th>\n",
              "      <th>1465</th>\n",
              "      <th>808</th>\n",
              "      <th>5792</th>\n",
              "      <th>3337</th>\n",
              "      <th>7494</th>\n",
              "      <th>2896</th>\n",
              "      <th>3098</th>\n",
              "      <th>11014</th>\n",
              "      <th>6500</th>\n",
              "      <th>823</th>\n",
              "      <th>813</th>\n",
              "      <th>1509</th>\n",
              "      <th>4144</th>\n",
              "      <th>3925</th>\n",
              "      <th>1891</th>\n",
              "      <th>8826</th>\n",
              "      <th>3303</th>\n",
              "      <th>3329</th>\n",
              "      <th>1153</th>\n",
              "      <th>5720</th>\n",
              "      <th>5048</th>\n",
              "      <th>10525</th>\n",
              "      <th>5708</th>\n",
              "      <th>2058</th>\n",
              "      <th>5710</th>\n",
              "      <th>5223</th>\n",
              "      <th>...</th>\n",
              "      <th>11182</th>\n",
              "      <th>63933</th>\n",
              "      <th>54850</th>\n",
              "      <th>29916</th>\n",
              "      <th>55179</th>\n",
              "      <th>29083</th>\n",
              "      <th>7099</th>\n",
              "      <th>79961</th>\n",
              "      <th>55825</th>\n",
              "      <th>51742</th>\n",
              "      <th>10818</th>\n",
              "      <th>81533</th>\n",
              "      <th>665</th>\n",
              "      <th>10776</th>\n",
              "      <th>64422</th>\n",
              "      <th>10813</th>\n",
              "      <th>80349</th>\n",
              "      <th>1978</th>\n",
              "      <th>83743</th>\n",
              "      <th>57406</th>\n",
              "      <th>23410</th>\n",
              "      <th>8996</th>\n",
              "      <th>55620</th>\n",
              "      <th>23597</th>\n",
              "      <th>124583</th>\n",
              "      <th>54915</th>\n",
              "      <th>3157</th>\n",
              "      <th>85377</th>\n",
              "      <th>55793</th>\n",
              "      <th>466</th>\n",
              "      <th>54681</th>\n",
              "      <th>11000</th>\n",
              "      <th>6915</th>\n",
              "      <th>6253</th>\n",
              "      <th>7264</th>\n",
              "      <th>5467</th>\n",
              "      <th>2767</th>\n",
              "      <th>23038</th>\n",
              "      <th>57048</th>\n",
              "      <th>79716</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cid</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:A03</th>\n",
              "      <td>4.264143</td>\n",
              "      <td>0.057249</td>\n",
              "      <td>-2.139334</td>\n",
              "      <td>-0.221784</td>\n",
              "      <td>-0.376555</td>\n",
              "      <td>1.913788</td>\n",
              "      <td>0.065576</td>\n",
              "      <td>-0.868953</td>\n",
              "      <td>-0.102273</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.525076</td>\n",
              "      <td>0.494065</td>\n",
              "      <td>-0.179588</td>\n",
              "      <td>-0.617669</td>\n",
              "      <td>-0.058411</td>\n",
              "      <td>0.247190</td>\n",
              "      <td>-0.432858</td>\n",
              "      <td>-0.321727</td>\n",
              "      <td>-1.285001</td>\n",
              "      <td>-0.388950</td>\n",
              "      <td>-0.088958</td>\n",
              "      <td>-0.097954</td>\n",
              "      <td>0.685952</td>\n",
              "      <td>0.959470</td>\n",
              "      <td>0.559862</td>\n",
              "      <td>-2.502630</td>\n",
              "      <td>0.359842</td>\n",
              "      <td>0.217574</td>\n",
              "      <td>-0.128369</td>\n",
              "      <td>-0.481144</td>\n",
              "      <td>2.042193</td>\n",
              "      <td>0.135226</td>\n",
              "      <td>0.144666</td>\n",
              "      <td>0.387548</td>\n",
              "      <td>0.729601</td>\n",
              "      <td>-0.751490</td>\n",
              "      <td>0.224599</td>\n",
              "      <td>0.833347</td>\n",
              "      <td>-0.174646</td>\n",
              "      <td>-0.300092</td>\n",
              "      <td>...</td>\n",
              "      <td>1.161566</td>\n",
              "      <td>0.096894</td>\n",
              "      <td>-0.609974</td>\n",
              "      <td>0.984338</td>\n",
              "      <td>1.298751</td>\n",
              "      <td>-0.444139</td>\n",
              "      <td>-0.077251</td>\n",
              "      <td>-0.863196</td>\n",
              "      <td>-2.093450</td>\n",
              "      <td>0.892946</td>\n",
              "      <td>-0.151042</td>\n",
              "      <td>0.596993</td>\n",
              "      <td>-0.246462</td>\n",
              "      <td>0.478527</td>\n",
              "      <td>0.458574</td>\n",
              "      <td>-0.213532</td>\n",
              "      <td>1.343217</td>\n",
              "      <td>-0.918547</td>\n",
              "      <td>-0.231647</td>\n",
              "      <td>-0.885711</td>\n",
              "      <td>4.704529</td>\n",
              "      <td>-2.583687</td>\n",
              "      <td>-0.847691</td>\n",
              "      <td>-0.324260</td>\n",
              "      <td>0.621623</td>\n",
              "      <td>-0.122100</td>\n",
              "      <td>-0.189515</td>\n",
              "      <td>0.455307</td>\n",
              "      <td>-0.686222</td>\n",
              "      <td>0.886410</td>\n",
              "      <td>5.246200</td>\n",
              "      <td>0.057058</td>\n",
              "      <td>-1.256173</td>\n",
              "      <td>-0.025255</td>\n",
              "      <td>0.168856</td>\n",
              "      <td>-1.639110</td>\n",
              "      <td>0.685100</td>\n",
              "      <td>-0.419421</td>\n",
              "      <td>1.716090</td>\n",
              "      <td>-0.472335</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:A05</th>\n",
              "      <td>-0.571711</td>\n",
              "      <td>-0.754999</td>\n",
              "      <td>-0.710110</td>\n",
              "      <td>0.428894</td>\n",
              "      <td>0.272606</td>\n",
              "      <td>0.314359</td>\n",
              "      <td>-1.545846</td>\n",
              "      <td>0.641955</td>\n",
              "      <td>-0.876852</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.690696</td>\n",
              "      <td>0.259465</td>\n",
              "      <td>0.097236</td>\n",
              "      <td>0.346196</td>\n",
              "      <td>-3.480344</td>\n",
              "      <td>1.192023</td>\n",
              "      <td>-0.129932</td>\n",
              "      <td>-0.620908</td>\n",
              "      <td>-0.145761</td>\n",
              "      <td>-0.332579</td>\n",
              "      <td>-0.397863</td>\n",
              "      <td>-0.064081</td>\n",
              "      <td>-0.068735</td>\n",
              "      <td>-0.371589</td>\n",
              "      <td>0.399846</td>\n",
              "      <td>-0.188653</td>\n",
              "      <td>0.410684</td>\n",
              "      <td>0.174939</td>\n",
              "      <td>0.144009</td>\n",
              "      <td>-0.855949</td>\n",
              "      <td>0.419143</td>\n",
              "      <td>0.437832</td>\n",
              "      <td>-0.056314</td>\n",
              "      <td>-0.193758</td>\n",
              "      <td>0.309855</td>\n",
              "      <td>-0.107573</td>\n",
              "      <td>-0.274771</td>\n",
              "      <td>0.739789</td>\n",
              "      <td>0.168686</td>\n",
              "      <td>-0.276458</td>\n",
              "      <td>...</td>\n",
              "      <td>0.951568</td>\n",
              "      <td>0.153254</td>\n",
              "      <td>0.333606</td>\n",
              "      <td>-1.179648</td>\n",
              "      <td>-0.143009</td>\n",
              "      <td>-3.184037</td>\n",
              "      <td>0.056141</td>\n",
              "      <td>-0.642681</td>\n",
              "      <td>-0.075741</td>\n",
              "      <td>0.566838</td>\n",
              "      <td>-0.080453</td>\n",
              "      <td>0.465662</td>\n",
              "      <td>-1.577573</td>\n",
              "      <td>0.348894</td>\n",
              "      <td>0.720928</td>\n",
              "      <td>0.386417</td>\n",
              "      <td>-0.043855</td>\n",
              "      <td>-0.482106</td>\n",
              "      <td>-0.964702</td>\n",
              "      <td>-0.561778</td>\n",
              "      <td>3.391710</td>\n",
              "      <td>-0.009394</td>\n",
              "      <td>0.393942</td>\n",
              "      <td>-0.046790</td>\n",
              "      <td>0.572470</td>\n",
              "      <td>1.012875</td>\n",
              "      <td>-0.393289</td>\n",
              "      <td>0.280188</td>\n",
              "      <td>0.407029</td>\n",
              "      <td>-0.021540</td>\n",
              "      <td>-0.005612</td>\n",
              "      <td>1.194516</td>\n",
              "      <td>0.171107</td>\n",
              "      <td>0.334370</td>\n",
              "      <td>-1.089265</td>\n",
              "      <td>-0.114638</td>\n",
              "      <td>0.304832</td>\n",
              "      <td>-0.249467</td>\n",
              "      <td>-0.428352</td>\n",
              "      <td>1.584976</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:B06</th>\n",
              "      <td>0.090035</td>\n",
              "      <td>-0.126354</td>\n",
              "      <td>-0.677804</td>\n",
              "      <td>0.018699</td>\n",
              "      <td>-0.171359</td>\n",
              "      <td>-0.929230</td>\n",
              "      <td>-0.518167</td>\n",
              "      <td>0.909588</td>\n",
              "      <td>-0.629458</td>\n",
              "      <td>0.0</td>\n",
              "      <td>-0.634575</td>\n",
              "      <td>-0.890770</td>\n",
              "      <td>0.184573</td>\n",
              "      <td>-0.052825</td>\n",
              "      <td>0.424141</td>\n",
              "      <td>0.070593</td>\n",
              "      <td>0.385842</td>\n",
              "      <td>-0.593042</td>\n",
              "      <td>0.129773</td>\n",
              "      <td>-0.033163</td>\n",
              "      <td>0.311473</td>\n",
              "      <td>-0.330706</td>\n",
              "      <td>0.164594</td>\n",
              "      <td>-0.221457</td>\n",
              "      <td>-0.005973</td>\n",
              "      <td>-0.995336</td>\n",
              "      <td>0.120785</td>\n",
              "      <td>0.053835</td>\n",
              "      <td>0.398550</td>\n",
              "      <td>-0.883482</td>\n",
              "      <td>0.685645</td>\n",
              "      <td>-0.013409</td>\n",
              "      <td>0.732580</td>\n",
              "      <td>-0.587955</td>\n",
              "      <td>0.539009</td>\n",
              "      <td>-0.578586</td>\n",
              "      <td>1.014374</td>\n",
              "      <td>0.081978</td>\n",
              "      <td>0.507190</td>\n",
              "      <td>-0.581820</td>\n",
              "      <td>...</td>\n",
              "      <td>0.413326</td>\n",
              "      <td>1.103752</td>\n",
              "      <td>0.282647</td>\n",
              "      <td>0.147169</td>\n",
              "      <td>0.148795</td>\n",
              "      <td>-0.378066</td>\n",
              "      <td>-0.546149</td>\n",
              "      <td>-0.626941</td>\n",
              "      <td>-0.330729</td>\n",
              "      <td>-0.066034</td>\n",
              "      <td>0.085613</td>\n",
              "      <td>-0.283772</td>\n",
              "      <td>-0.040362</td>\n",
              "      <td>0.326949</td>\n",
              "      <td>0.807251</td>\n",
              "      <td>-0.090438</td>\n",
              "      <td>-0.043841</td>\n",
              "      <td>-0.120055</td>\n",
              "      <td>-0.403158</td>\n",
              "      <td>-0.087850</td>\n",
              "      <td>-0.310649</td>\n",
              "      <td>-0.350287</td>\n",
              "      <td>-0.580429</td>\n",
              "      <td>-0.122886</td>\n",
              "      <td>0.505836</td>\n",
              "      <td>-0.156736</td>\n",
              "      <td>-0.225532</td>\n",
              "      <td>-0.142011</td>\n",
              "      <td>4.102704</td>\n",
              "      <td>-0.957159</td>\n",
              "      <td>0.084765</td>\n",
              "      <td>4.083964</td>\n",
              "      <td>0.736748</td>\n",
              "      <td>-0.130357</td>\n",
              "      <td>0.229097</td>\n",
              "      <td>0.445147</td>\n",
              "      <td>0.391595</td>\n",
              "      <td>0.203354</td>\n",
              "      <td>-0.038053</td>\n",
              "      <td>1.808435</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:F08</th>\n",
              "      <td>0.486323</td>\n",
              "      <td>-0.048152</td>\n",
              "      <td>-0.026651</td>\n",
              "      <td>-0.189330</td>\n",
              "      <td>0.746427</td>\n",
              "      <td>0.543150</td>\n",
              "      <td>-0.041739</td>\n",
              "      <td>-0.155129</td>\n",
              "      <td>-0.080053</td>\n",
              "      <td>0.0</td>\n",
              "      <td>-0.231303</td>\n",
              "      <td>-0.350301</td>\n",
              "      <td>-0.478873</td>\n",
              "      <td>-0.303196</td>\n",
              "      <td>-1.122758</td>\n",
              "      <td>-0.690514</td>\n",
              "      <td>-0.402801</td>\n",
              "      <td>-0.385334</td>\n",
              "      <td>-0.081466</td>\n",
              "      <td>0.572392</td>\n",
              "      <td>0.531504</td>\n",
              "      <td>-0.543804</td>\n",
              "      <td>0.836444</td>\n",
              "      <td>0.302225</td>\n",
              "      <td>0.624687</td>\n",
              "      <td>0.161752</td>\n",
              "      <td>0.196965</td>\n",
              "      <td>-0.036358</td>\n",
              "      <td>-0.683191</td>\n",
              "      <td>-0.037815</td>\n",
              "      <td>-3.346440</td>\n",
              "      <td>0.127267</td>\n",
              "      <td>-0.331009</td>\n",
              "      <td>0.524529</td>\n",
              "      <td>-0.330875</td>\n",
              "      <td>-0.063817</td>\n",
              "      <td>-0.292505</td>\n",
              "      <td>0.148976</td>\n",
              "      <td>0.989419</td>\n",
              "      <td>0.475239</td>\n",
              "      <td>...</td>\n",
              "      <td>0.278010</td>\n",
              "      <td>0.889349</td>\n",
              "      <td>0.194808</td>\n",
              "      <td>0.356561</td>\n",
              "      <td>-0.019010</td>\n",
              "      <td>-0.062983</td>\n",
              "      <td>0.317133</td>\n",
              "      <td>0.100935</td>\n",
              "      <td>0.297193</td>\n",
              "      <td>0.496246</td>\n",
              "      <td>0.427417</td>\n",
              "      <td>-0.396475</td>\n",
              "      <td>0.593992</td>\n",
              "      <td>-2.287117</td>\n",
              "      <td>0.214099</td>\n",
              "      <td>-0.154662</td>\n",
              "      <td>0.111357</td>\n",
              "      <td>-0.589143</td>\n",
              "      <td>-0.318355</td>\n",
              "      <td>0.214677</td>\n",
              "      <td>-0.138251</td>\n",
              "      <td>0.085525</td>\n",
              "      <td>-0.622205</td>\n",
              "      <td>-1.362202</td>\n",
              "      <td>-0.138776</td>\n",
              "      <td>1.148892</td>\n",
              "      <td>0.809686</td>\n",
              "      <td>-0.360932</td>\n",
              "      <td>0.367233</td>\n",
              "      <td>-0.328657</td>\n",
              "      <td>0.051168</td>\n",
              "      <td>-0.141523</td>\n",
              "      <td>0.750948</td>\n",
              "      <td>-0.458847</td>\n",
              "      <td>-0.501552</td>\n",
              "      <td>0.674004</td>\n",
              "      <td>-0.091482</td>\n",
              "      <td>0.245796</td>\n",
              "      <td>0.853410</td>\n",
              "      <td>-0.368149</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:F11</th>\n",
              "      <td>0.782842</td>\n",
              "      <td>0.023534</td>\n",
              "      <td>0.295156</td>\n",
              "      <td>-0.415375</td>\n",
              "      <td>1.151106</td>\n",
              "      <td>0.086564</td>\n",
              "      <td>0.271207</td>\n",
              "      <td>-0.450202</td>\n",
              "      <td>0.494099</td>\n",
              "      <td>0.0</td>\n",
              "      <td>0.069646</td>\n",
              "      <td>-0.189121</td>\n",
              "      <td>0.259846</td>\n",
              "      <td>0.307957</td>\n",
              "      <td>0.206965</td>\n",
              "      <td>0.182580</td>\n",
              "      <td>-1.241487</td>\n",
              "      <td>0.166138</td>\n",
              "      <td>0.047024</td>\n",
              "      <td>0.351032</td>\n",
              "      <td>0.186668</td>\n",
              "      <td>0.499309</td>\n",
              "      <td>0.678594</td>\n",
              "      <td>-0.656629</td>\n",
              "      <td>-0.135447</td>\n",
              "      <td>-1.372328</td>\n",
              "      <td>-0.038117</td>\n",
              "      <td>0.261096</td>\n",
              "      <td>-0.367883</td>\n",
              "      <td>0.803902</td>\n",
              "      <td>-0.768213</td>\n",
              "      <td>0.639764</td>\n",
              "      <td>-0.389107</td>\n",
              "      <td>1.236705</td>\n",
              "      <td>0.028638</td>\n",
              "      <td>4.135058</td>\n",
              "      <td>-0.353838</td>\n",
              "      <td>-0.462894</td>\n",
              "      <td>1.140511</td>\n",
              "      <td>0.831206</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.256794</td>\n",
              "      <td>1.249464</td>\n",
              "      <td>-0.756335</td>\n",
              "      <td>1.062800</td>\n",
              "      <td>0.008821</td>\n",
              "      <td>-0.008758</td>\n",
              "      <td>-0.136715</td>\n",
              "      <td>-0.684813</td>\n",
              "      <td>0.209824</td>\n",
              "      <td>-0.489167</td>\n",
              "      <td>0.263006</td>\n",
              "      <td>-0.227860</td>\n",
              "      <td>0.714842</td>\n",
              "      <td>0.751042</td>\n",
              "      <td>0.394874</td>\n",
              "      <td>0.638209</td>\n",
              "      <td>0.432576</td>\n",
              "      <td>-0.598256</td>\n",
              "      <td>0.050784</td>\n",
              "      <td>1.414118</td>\n",
              "      <td>0.229658</td>\n",
              "      <td>-0.150874</td>\n",
              "      <td>-0.674186</td>\n",
              "      <td>-0.466741</td>\n",
              "      <td>-0.051128</td>\n",
              "      <td>-0.528998</td>\n",
              "      <td>0.313247</td>\n",
              "      <td>0.275544</td>\n",
              "      <td>0.984492</td>\n",
              "      <td>0.222858</td>\n",
              "      <td>0.524670</td>\n",
              "      <td>-0.574846</td>\n",
              "      <td>-0.420265</td>\n",
              "      <td>0.254262</td>\n",
              "      <td>0.095857</td>\n",
              "      <td>0.947532</td>\n",
              "      <td>0.535357</td>\n",
              "      <td>-0.018437</td>\n",
              "      <td>-0.720413</td>\n",
              "      <td>0.065324</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 978 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                            780      7849  ...     57048     79716\n",
              "cid                                        ...                    \n",
              "REP.A001_A375_24H:A03  4.264143  0.057249  ...  1.716090 -0.472335\n",
              "REP.A001_A375_24H:A05 -0.571711 -0.754999  ... -0.428352  1.584976\n",
              "REP.A001_A375_24H:B06  0.090035 -0.126354  ... -0.038053  1.808435\n",
              "REP.A001_A375_24H:F08  0.486323 -0.048152  ...  0.853410 -0.368149\n",
              "REP.A001_A375_24H:F11  0.782842  0.023534  ... -0.720413  0.065324\n",
              "\n",
              "[5 rows x 978 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 26
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JPEq33WRg_r-",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 284
        },
        "outputId": "434bcdc8-8b39-4042-bda7-e8a3b0daf43a"
      },
      "source": [
        "DDRi_gene_expr_rn=random_noise(DDRi_gene_expr)\n",
        "DDRi_gene_expr_rn.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>780</th>\n",
              "      <th>7849</th>\n",
              "      <th>6193</th>\n",
              "      <th>23</th>\n",
              "      <th>9552</th>\n",
              "      <th>387</th>\n",
              "      <th>10921</th>\n",
              "      <th>10285</th>\n",
              "      <th>533</th>\n",
              "      <th>6194</th>\n",
              "      <th>23326</th>\n",
              "      <th>351</th>\n",
              "      <th>1213</th>\n",
              "      <th>9761</th>\n",
              "      <th>1465</th>\n",
              "      <th>808</th>\n",
              "      <th>5792</th>\n",
              "      <th>3337</th>\n",
              "      <th>7494</th>\n",
              "      <th>2896</th>\n",
              "      <th>3098</th>\n",
              "      <th>11014</th>\n",
              "      <th>6500</th>\n",
              "      <th>823</th>\n",
              "      <th>813</th>\n",
              "      <th>1509</th>\n",
              "      <th>4144</th>\n",
              "      <th>3925</th>\n",
              "      <th>1891</th>\n",
              "      <th>8826</th>\n",
              "      <th>3303</th>\n",
              "      <th>3329</th>\n",
              "      <th>1153</th>\n",
              "      <th>5720</th>\n",
              "      <th>5048</th>\n",
              "      <th>10525</th>\n",
              "      <th>5708</th>\n",
              "      <th>2058</th>\n",
              "      <th>5710</th>\n",
              "      <th>5223</th>\n",
              "      <th>...</th>\n",
              "      <th>11182</th>\n",
              "      <th>63933</th>\n",
              "      <th>54850</th>\n",
              "      <th>29916</th>\n",
              "      <th>55179</th>\n",
              "      <th>29083</th>\n",
              "      <th>7099</th>\n",
              "      <th>79961</th>\n",
              "      <th>55825</th>\n",
              "      <th>51742</th>\n",
              "      <th>10818</th>\n",
              "      <th>81533</th>\n",
              "      <th>665</th>\n",
              "      <th>10776</th>\n",
              "      <th>64422</th>\n",
              "      <th>10813</th>\n",
              "      <th>80349</th>\n",
              "      <th>1978</th>\n",
              "      <th>83743</th>\n",
              "      <th>57406</th>\n",
              "      <th>23410</th>\n",
              "      <th>8996</th>\n",
              "      <th>55620</th>\n",
              "      <th>23597</th>\n",
              "      <th>124583</th>\n",
              "      <th>54915</th>\n",
              "      <th>3157</th>\n",
              "      <th>85377</th>\n",
              "      <th>55793</th>\n",
              "      <th>466</th>\n",
              "      <th>54681</th>\n",
              "      <th>11000</th>\n",
              "      <th>6915</th>\n",
              "      <th>6253</th>\n",
              "      <th>7264</th>\n",
              "      <th>5467</th>\n",
              "      <th>2767</th>\n",
              "      <th>23038</th>\n",
              "      <th>57048</th>\n",
              "      <th>79716</th>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>cid</th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "      <th></th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:A03</th>\n",
              "      <td>0.346951</td>\n",
              "      <td>-0.072225</td>\n",
              "      <td>0.263486</td>\n",
              "      <td>-0.911088</td>\n",
              "      <td>0.034348</td>\n",
              "      <td>0.133839</td>\n",
              "      <td>-0.465517</td>\n",
              "      <td>0.699463</td>\n",
              "      <td>0.179825</td>\n",
              "      <td>0.000000</td>\n",
              "      <td>-0.397597</td>\n",
              "      <td>-0.935994</td>\n",
              "      <td>-0.020811</td>\n",
              "      <td>-1.028512</td>\n",
              "      <td>-2.519510</td>\n",
              "      <td>0.254389</td>\n",
              "      <td>-0.121619</td>\n",
              "      <td>1.249585</td>\n",
              "      <td>0.117861</td>\n",
              "      <td>-0.831263</td>\n",
              "      <td>-0.771881</td>\n",
              "      <td>-1.938656</td>\n",
              "      <td>-0.007787</td>\n",
              "      <td>-0.585267</td>\n",
              "      <td>0.232465</td>\n",
              "      <td>-0.456491</td>\n",
              "      <td>0.447082</td>\n",
              "      <td>0.337116</td>\n",
              "      <td>-0.180163</td>\n",
              "      <td>-0.973887</td>\n",
              "      <td>-0.019860</td>\n",
              "      <td>0.731367</td>\n",
              "      <td>-0.036499</td>\n",
              "      <td>0.566481</td>\n",
              "      <td>-0.667920</td>\n",
              "      <td>0.513303</td>\n",
              "      <td>-0.074567</td>\n",
              "      <td>-0.942845</td>\n",
              "      <td>0.096436</td>\n",
              "      <td>-0.087692</td>\n",
              "      <td>...</td>\n",
              "      <td>0.164731</td>\n",
              "      <td>1.824395</td>\n",
              "      <td>-0.001707</td>\n",
              "      <td>2.955862</td>\n",
              "      <td>0.305478</td>\n",
              "      <td>0.209582</td>\n",
              "      <td>1.848617</td>\n",
              "      <td>1.012357</td>\n",
              "      <td>0.696036</td>\n",
              "      <td>-0.684889</td>\n",
              "      <td>-0.582877</td>\n",
              "      <td>-1.399600</td>\n",
              "      <td>-0.889834</td>\n",
              "      <td>-1.178846</td>\n",
              "      <td>0.279318</td>\n",
              "      <td>1.119270</td>\n",
              "      <td>0.407685</td>\n",
              "      <td>1.299560</td>\n",
              "      <td>-2.740435</td>\n",
              "      <td>-0.693942</td>\n",
              "      <td>0.892831</td>\n",
              "      <td>0.026476</td>\n",
              "      <td>-0.085264</td>\n",
              "      <td>0.417326</td>\n",
              "      <td>-0.639990</td>\n",
              "      <td>0.215025</td>\n",
              "      <td>-0.199806</td>\n",
              "      <td>-0.572732</td>\n",
              "      <td>0.147290</td>\n",
              "      <td>0.025148</td>\n",
              "      <td>0.203578</td>\n",
              "      <td>0.950807</td>\n",
              "      <td>3.208275</td>\n",
              "      <td>1.185582</td>\n",
              "      <td>0.460981</td>\n",
              "      <td>-1.106293</td>\n",
              "      <td>0.601467</td>\n",
              "      <td>0.757136</td>\n",
              "      <td>0.427717</td>\n",
              "      <td>-1.037486</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:A05</th>\n",
              "      <td>0.583586</td>\n",
              "      <td>-1.313570</td>\n",
              "      <td>-0.054442</td>\n",
              "      <td>-2.159236</td>\n",
              "      <td>-0.344172</td>\n",
              "      <td>1.239863</td>\n",
              "      <td>-0.336611</td>\n",
              "      <td>-0.919846</td>\n",
              "      <td>0.898237</td>\n",
              "      <td>0.001418</td>\n",
              "      <td>0.674341</td>\n",
              "      <td>0.000241</td>\n",
              "      <td>1.134536</td>\n",
              "      <td>0.930300</td>\n",
              "      <td>-0.306106</td>\n",
              "      <td>-2.298258</td>\n",
              "      <td>-2.097048</td>\n",
              "      <td>0.406541</td>\n",
              "      <td>0.259399</td>\n",
              "      <td>-0.114275</td>\n",
              "      <td>0.733237</td>\n",
              "      <td>0.975634</td>\n",
              "      <td>0.335389</td>\n",
              "      <td>0.499952</td>\n",
              "      <td>0.256970</td>\n",
              "      <td>-1.428313</td>\n",
              "      <td>1.880708</td>\n",
              "      <td>1.566421</td>\n",
              "      <td>-1.225143</td>\n",
              "      <td>-2.058756</td>\n",
              "      <td>-0.678872</td>\n",
              "      <td>-0.331356</td>\n",
              "      <td>0.468274</td>\n",
              "      <td>0.260269</td>\n",
              "      <td>0.187290</td>\n",
              "      <td>-1.597586</td>\n",
              "      <td>-1.005662</td>\n",
              "      <td>0.471443</td>\n",
              "      <td>0.460908</td>\n",
              "      <td>-0.443206</td>\n",
              "      <td>...</td>\n",
              "      <td>0.882257</td>\n",
              "      <td>1.213797</td>\n",
              "      <td>2.098609</td>\n",
              "      <td>1.041858</td>\n",
              "      <td>-1.285834</td>\n",
              "      <td>-0.584476</td>\n",
              "      <td>-0.707017</td>\n",
              "      <td>0.173405</td>\n",
              "      <td>-1.090760</td>\n",
              "      <td>2.229658</td>\n",
              "      <td>-0.019146</td>\n",
              "      <td>-0.681150</td>\n",
              "      <td>1.138265</td>\n",
              "      <td>1.025823</td>\n",
              "      <td>-0.144150</td>\n",
              "      <td>-1.000786</td>\n",
              "      <td>0.623558</td>\n",
              "      <td>0.022526</td>\n",
              "      <td>0.956489</td>\n",
              "      <td>-0.792409</td>\n",
              "      <td>3.127322</td>\n",
              "      <td>-1.181600</td>\n",
              "      <td>-1.138221</td>\n",
              "      <td>-0.350499</td>\n",
              "      <td>0.295302</td>\n",
              "      <td>0.165968</td>\n",
              "      <td>0.000991</td>\n",
              "      <td>4.224746</td>\n",
              "      <td>-0.318300</td>\n",
              "      <td>1.370810</td>\n",
              "      <td>0.407046</td>\n",
              "      <td>-0.924363</td>\n",
              "      <td>-1.165396</td>\n",
              "      <td>-0.933684</td>\n",
              "      <td>0.191026</td>\n",
              "      <td>-0.790862</td>\n",
              "      <td>0.696797</td>\n",
              "      <td>1.429895</td>\n",
              "      <td>1.126931</td>\n",
              "      <td>-0.826261</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:B06</th>\n",
              "      <td>-0.239851</td>\n",
              "      <td>-0.513102</td>\n",
              "      <td>-0.777243</td>\n",
              "      <td>-0.314719</td>\n",
              "      <td>-0.589689</td>\n",
              "      <td>0.644182</td>\n",
              "      <td>0.190117</td>\n",
              "      <td>-0.473368</td>\n",
              "      <td>-0.155920</td>\n",
              "      <td>0.484108</td>\n",
              "      <td>1.252349</td>\n",
              "      <td>-0.757016</td>\n",
              "      <td>0.284382</td>\n",
              "      <td>-0.903291</td>\n",
              "      <td>0.152020</td>\n",
              "      <td>-0.132840</td>\n",
              "      <td>-1.049063</td>\n",
              "      <td>0.261602</td>\n",
              "      <td>0.122313</td>\n",
              "      <td>0.201640</td>\n",
              "      <td>0.503672</td>\n",
              "      <td>0.546444</td>\n",
              "      <td>0.202881</td>\n",
              "      <td>0.129837</td>\n",
              "      <td>0.378768</td>\n",
              "      <td>0.674284</td>\n",
              "      <td>0.281716</td>\n",
              "      <td>0.242391</td>\n",
              "      <td>-0.462490</td>\n",
              "      <td>0.147127</td>\n",
              "      <td>0.367566</td>\n",
              "      <td>0.728866</td>\n",
              "      <td>0.904291</td>\n",
              "      <td>0.170927</td>\n",
              "      <td>-0.275767</td>\n",
              "      <td>0.045164</td>\n",
              "      <td>0.345351</td>\n",
              "      <td>-0.182215</td>\n",
              "      <td>0.398809</td>\n",
              "      <td>0.458738</td>\n",
              "      <td>...</td>\n",
              "      <td>-0.235727</td>\n",
              "      <td>1.571898</td>\n",
              "      <td>-0.781109</td>\n",
              "      <td>1.323763</td>\n",
              "      <td>-0.319975</td>\n",
              "      <td>-0.346691</td>\n",
              "      <td>3.509586</td>\n",
              "      <td>0.263932</td>\n",
              "      <td>2.286774</td>\n",
              "      <td>-0.156389</td>\n",
              "      <td>-0.010751</td>\n",
              "      <td>-1.691814</td>\n",
              "      <td>-0.560015</td>\n",
              "      <td>0.506985</td>\n",
              "      <td>0.753619</td>\n",
              "      <td>-3.069787</td>\n",
              "      <td>0.164010</td>\n",
              "      <td>0.554008</td>\n",
              "      <td>0.759607</td>\n",
              "      <td>-0.393747</td>\n",
              "      <td>2.755333</td>\n",
              "      <td>0.559180</td>\n",
              "      <td>0.133438</td>\n",
              "      <td>-0.678692</td>\n",
              "      <td>0.395368</td>\n",
              "      <td>0.575078</td>\n",
              "      <td>-0.044171</td>\n",
              "      <td>0.488836</td>\n",
              "      <td>0.427758</td>\n",
              "      <td>0.236872</td>\n",
              "      <td>0.207849</td>\n",
              "      <td>-0.343398</td>\n",
              "      <td>-0.633409</td>\n",
              "      <td>-1.247897</td>\n",
              "      <td>-0.711078</td>\n",
              "      <td>-0.215181</td>\n",
              "      <td>0.225914</td>\n",
              "      <td>0.026922</td>\n",
              "      <td>-1.650863</td>\n",
              "      <td>0.638502</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:F08</th>\n",
              "      <td>-0.766225</td>\n",
              "      <td>-1.275173</td>\n",
              "      <td>0.679984</td>\n",
              "      <td>-0.253703</td>\n",
              "      <td>-1.927102</td>\n",
              "      <td>0.515917</td>\n",
              "      <td>1.585518</td>\n",
              "      <td>1.094815</td>\n",
              "      <td>0.499072</td>\n",
              "      <td>0.028714</td>\n",
              "      <td>0.427401</td>\n",
              "      <td>-0.981079</td>\n",
              "      <td>0.477132</td>\n",
              "      <td>-0.045834</td>\n",
              "      <td>0.836801</td>\n",
              "      <td>-0.210036</td>\n",
              "      <td>-0.690459</td>\n",
              "      <td>0.250163</td>\n",
              "      <td>0.137765</td>\n",
              "      <td>-0.597090</td>\n",
              "      <td>0.006441</td>\n",
              "      <td>0.649895</td>\n",
              "      <td>0.823322</td>\n",
              "      <td>-0.245850</td>\n",
              "      <td>-0.145002</td>\n",
              "      <td>-0.847882</td>\n",
              "      <td>0.644175</td>\n",
              "      <td>-0.069473</td>\n",
              "      <td>-1.531104</td>\n",
              "      <td>0.002089</td>\n",
              "      <td>0.020892</td>\n",
              "      <td>0.674786</td>\n",
              "      <td>-0.824802</td>\n",
              "      <td>-0.386345</td>\n",
              "      <td>-0.734920</td>\n",
              "      <td>-1.200396</td>\n",
              "      <td>-0.591726</td>\n",
              "      <td>-0.842944</td>\n",
              "      <td>0.349293</td>\n",
              "      <td>0.242944</td>\n",
              "      <td>...</td>\n",
              "      <td>-1.122287</td>\n",
              "      <td>0.128276</td>\n",
              "      <td>-0.140897</td>\n",
              "      <td>0.774912</td>\n",
              "      <td>-0.790092</td>\n",
              "      <td>0.992719</td>\n",
              "      <td>0.459722</td>\n",
              "      <td>-0.815332</td>\n",
              "      <td>-0.532001</td>\n",
              "      <td>-0.222669</td>\n",
              "      <td>0.093182</td>\n",
              "      <td>-0.893619</td>\n",
              "      <td>0.565390</td>\n",
              "      <td>0.352403</td>\n",
              "      <td>-1.029945</td>\n",
              "      <td>0.534839</td>\n",
              "      <td>1.441194</td>\n",
              "      <td>0.545300</td>\n",
              "      <td>0.932863</td>\n",
              "      <td>-0.880891</td>\n",
              "      <td>-0.206184</td>\n",
              "      <td>-0.912651</td>\n",
              "      <td>-0.653018</td>\n",
              "      <td>1.017586</td>\n",
              "      <td>-2.933846</td>\n",
              "      <td>0.840421</td>\n",
              "      <td>0.054654</td>\n",
              "      <td>0.357397</td>\n",
              "      <td>-0.531720</td>\n",
              "      <td>0.998522</td>\n",
              "      <td>-0.822373</td>\n",
              "      <td>-1.183868</td>\n",
              "      <td>1.491937</td>\n",
              "      <td>-1.063638</td>\n",
              "      <td>0.629134</td>\n",
              "      <td>0.367977</td>\n",
              "      <td>-0.542443</td>\n",
              "      <td>-0.246186</td>\n",
              "      <td>0.400520</td>\n",
              "      <td>-0.930262</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>REP.A001_A375_24H:F11</th>\n",
              "      <td>0.006897</td>\n",
              "      <td>-0.180128</td>\n",
              "      <td>-0.389726</td>\n",
              "      <td>0.550577</td>\n",
              "      <td>-0.759457</td>\n",
              "      <td>-0.279520</td>\n",
              "      <td>0.317637</td>\n",
              "      <td>0.560946</td>\n",
              "      <td>-0.264962</td>\n",
              "      <td>0.270989</td>\n",
              "      <td>-0.518505</td>\n",
              "      <td>-0.025443</td>\n",
              "      <td>0.165745</td>\n",
              "      <td>0.278006</td>\n",
              "      <td>-0.224238</td>\n",
              "      <td>-0.255243</td>\n",
              "      <td>-0.280360</td>\n",
              "      <td>0.107095</td>\n",
              "      <td>0.639675</td>\n",
              "      <td>0.165720</td>\n",
              "      <td>-0.013417</td>\n",
              "      <td>0.007822</td>\n",
              "      <td>0.261039</td>\n",
              "      <td>-0.991556</td>\n",
              "      <td>0.495247</td>\n",
              "      <td>0.435768</td>\n",
              "      <td>0.123973</td>\n",
              "      <td>0.076726</td>\n",
              "      <td>-0.276655</td>\n",
              "      <td>0.065011</td>\n",
              "      <td>-0.835046</td>\n",
              "      <td>-0.296712</td>\n",
              "      <td>0.137708</td>\n",
              "      <td>0.548029</td>\n",
              "      <td>-0.666521</td>\n",
              "      <td>-0.152734</td>\n",
              "      <td>-0.153368</td>\n",
              "      <td>-0.215102</td>\n",
              "      <td>0.510921</td>\n",
              "      <td>3.791181</td>\n",
              "      <td>...</td>\n",
              "      <td>0.258342</td>\n",
              "      <td>0.470314</td>\n",
              "      <td>0.254975</td>\n",
              "      <td>-0.104885</td>\n",
              "      <td>0.419373</td>\n",
              "      <td>0.138918</td>\n",
              "      <td>0.053057</td>\n",
              "      <td>0.840074</td>\n",
              "      <td>0.135155</td>\n",
              "      <td>-1.953834</td>\n",
              "      <td>0.370138</td>\n",
              "      <td>0.111195</td>\n",
              "      <td>-0.078811</td>\n",
              "      <td>3.625789</td>\n",
              "      <td>-0.062106</td>\n",
              "      <td>0.049769</td>\n",
              "      <td>1.047637</td>\n",
              "      <td>0.774778</td>\n",
              "      <td>0.697463</td>\n",
              "      <td>-0.627622</td>\n",
              "      <td>-0.118766</td>\n",
              "      <td>-0.186667</td>\n",
              "      <td>-0.800171</td>\n",
              "      <td>-3.309188</td>\n",
              "      <td>-3.682968</td>\n",
              "      <td>1.133365</td>\n",
              "      <td>0.254603</td>\n",
              "      <td>-0.350364</td>\n",
              "      <td>0.405546</td>\n",
              "      <td>-0.315409</td>\n",
              "      <td>0.990217</td>\n",
              "      <td>-0.348412</td>\n",
              "      <td>-0.297456</td>\n",
              "      <td>-0.066910</td>\n",
              "      <td>0.409134</td>\n",
              "      <td>-0.236175</td>\n",
              "      <td>-0.250800</td>\n",
              "      <td>0.431800</td>\n",
              "      <td>-1.921559</td>\n",
              "      <td>-1.077229</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "<p>5 rows × 978 columns</p>\n",
              "</div>"
            ],
            "text/plain": [
              "                            780      7849  ...     57048     79716\n",
              "cid                                        ...                    \n",
              "REP.A001_A375_24H:A03  0.346951 -0.072225  ...  0.427717 -1.037486\n",
              "REP.A001_A375_24H:A05  0.583586 -1.313570  ...  1.126931 -0.826261\n",
              "REP.A001_A375_24H:B06 -0.239851 -0.513102  ... -1.650863  0.638502\n",
              "REP.A001_A375_24H:F08 -0.766225 -1.275173  ...  0.400520 -0.930262\n",
              "REP.A001_A375_24H:F11  0.006897 -0.180128  ... -1.921559 -1.077229\n",
              "\n",
              "[5 rows x 978 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 31
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "v-ro2IeigYF-",
        "outputId": "f7f3e09f-2a67-43d4-ff85-1b7b9b1016dd"
      },
      "source": [
        "# define the dataframe to be used\n",
        "dataf = DDRi_gene_expr_rn\n",
        "# split datasets\n",
        "train_data, valid_data = train_valid_split(dataf)\n",
        "# initiate dataloader\n",
        "train_loader = dataloader(train_data, 80)\n",
        "valid_loader = dataloader(valid_data,80)"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "samples in training data: 2191 \n",
            "samples in validation data: 939\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Z1--luHGCtIm"
      },
      "source": [
        "Visualization of the model structure"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "eE9xD2VC4SSN",
        "outputId": "1bdeea6c-ef8c-4c7b-b17d-cef04ccb164b"
      },
      "source": [
        "!pip install torchviz"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Requirement already satisfied: torchviz in /usr/local/lib/python3.7/dist-packages (0.0.1)\n",
            "Requirement already satisfied: torch in /usr/local/lib/python3.7/dist-packages (from torchviz) (1.8.0+cu101)\n",
            "Requirement already satisfied: graphviz in /usr/local/lib/python3.7/dist-packages (from torchviz) (0.10.1)\n",
            "Requirement already satisfied: numpy in /usr/local/lib/python3.7/dist-packages (from torch->torchviz) (1.19.5)\n",
            "Requirement already satisfied: typing-extensions in /usr/local/lib/python3.7/dist-packages (from torch->torchviz) (3.7.4.3)\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 433
        },
        "id": "Hodqesax4RWA",
        "outputId": "270fbafc-f146-4dcd-8ab7-efc9b22d4017"
      },
      "source": [
        "# visualize the structure of the model\n",
        "from torchviz import make_dot, make_dot_from_trace\n",
        "x = next(train_loader)\n",
        "x = torch.tensor(x).float()\n",
        "make_dot(DA(x), params=dict(list(DA.named_parameters())))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<graphviz.dot.Digraph at 0x7f9b62ba41d0>"
            ],
            "image/svg+xml": "<?xml version=\"1.0\" encoding=\"UTF-8\" standalone=\"no\"?>\n<!DOCTYPE svg PUBLIC \"-//W3C//DTD SVG 1.1//EN\"\n \"http://www.w3.org/Graphics/SVG/1.1/DTD/svg11.dtd\">\n<!-- Generated by graphviz version 2.40.1 (20161225.0304)\n -->\n<!-- Title: %3 Pages: 1 -->\n<svg width=\"304pt\" height=\"309pt\"\n viewBox=\"0.00 0.00 304.00 309.00\" xmlns=\"http://www.w3.org/2000/svg\" xmlns:xlink=\"http://www.w3.org/1999/xlink\">\n<g id=\"graph0\" class=\"graph\" transform=\"scale(1 1) rotate(0) translate(4 305)\">\n<title>%3</title>\n<polygon fill=\"#ffffff\" stroke=\"transparent\" points=\"-4,4 -4,-305 300,-305 300,4 -4,4\"/>\n<!-- 140305286453520 -->\n<g id=\"node1\" class=\"node\">\n<title>140305286453520</title>\n<polygon fill=\"#caff70\" stroke=\"#000000\" points=\"192.5,-21 88.5,-21 88.5,0 192.5,0 192.5,-21\"/>\n<text text-anchor=\"middle\" x=\"140.5\" y=\"-7.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">AddmmBackward</text>\n</g>\n<!-- 140305286455120 -->\n<g id=\"node2\" class=\"node\">\n<title>140305286455120</title>\n<polygon fill=\"#add8e6\" stroke=\"#000000\" points=\"75,-91 0,-91 0,-57 75,-57 75,-91\"/>\n<text text-anchor=\"middle\" x=\"37.5\" y=\"-77.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">decoder.bias</text>\n<text text-anchor=\"middle\" x=\"37.5\" y=\"-64.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\"> (978)</text>\n</g>\n<!-- 140305286455120&#45;&gt;140305286453520 -->\n<g id=\"edge1\" class=\"edge\">\n<title>140305286455120&#45;&gt;140305286453520</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M65.1021,-56.9832C80.3999,-47.552 99.3124,-35.8924 114.3688,-26.61\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"116.5029,-29.4061 123.1784,-21.1788 112.8293,-23.4474 116.5029,-29.4061\"/>\n</g>\n<!-- 140303919525200 -->\n<g id=\"node3\" class=\"node\">\n<title>140303919525200</title>\n<polygon fill=\"#d3d3d3\" stroke=\"#000000\" points=\"187.5,-84.5 93.5,-84.5 93.5,-63.5 187.5,-63.5 187.5,-84.5\"/>\n<text text-anchor=\"middle\" x=\"140.5\" y=\"-70.9\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">ReluBackward1</text>\n</g>\n<!-- 140303919525200&#45;&gt;140305286453520 -->\n<g id=\"edge2\" class=\"edge\">\n<title>140303919525200&#45;&gt;140305286453520</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M140.5,-63.2281C140.5,-54.5091 140.5,-41.9699 140.5,-31.3068\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"144.0001,-31.1128 140.5,-21.1128 137.0001,-31.1129 144.0001,-31.1128\"/>\n</g>\n<!-- 140303919176080 -->\n<g id=\"node4\" class=\"node\">\n<title>140303919176080</title>\n<polygon fill=\"#d3d3d3\" stroke=\"#000000\" points=\"188.5,-154.5 84.5,-154.5 84.5,-133.5 188.5,-133.5 188.5,-154.5\"/>\n<text text-anchor=\"middle\" x=\"136.5\" y=\"-140.9\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">AddmmBackward</text>\n</g>\n<!-- 140303919176080&#45;&gt;140303919525200 -->\n<g id=\"edge3\" class=\"edge\">\n<title>140303919176080&#45;&gt;140303919525200</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M137.1075,-133.3685C137.689,-123.1925 138.5822,-107.5606 139.3062,-94.8912\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"142.8105,-94.915 139.8868,-84.7315 135.8219,-94.5156 142.8105,-94.915\"/>\n</g>\n<!-- 140303919176592 -->\n<g id=\"node5\" class=\"node\">\n<title>140303919176592</title>\n<polygon fill=\"#add8e6\" stroke=\"#000000\" points=\"130.5,-231 46.5,-231 46.5,-197 130.5,-197 130.5,-231\"/>\n<text text-anchor=\"middle\" x=\"88.5\" y=\"-217.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">encoder.0.bias</text>\n<text text-anchor=\"middle\" x=\"88.5\" y=\"-204.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\"> (50)</text>\n</g>\n<!-- 140303919176592&#45;&gt;140303919176080 -->\n<g id=\"edge4\" class=\"edge\">\n<title>140303919176592&#45;&gt;140303919176080</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M100.3652,-196.6966C107.3222,-186.551 116.1253,-173.7131 123.3008,-163.2488\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"126.3315,-165.0179 129.1003,-154.7913 120.5584,-161.0592 126.3315,-165.0179\"/>\n</g>\n<!-- 140303919175824 -->\n<g id=\"node6\" class=\"node\">\n<title>140303919175824</title>\n<polygon fill=\"#d3d3d3\" stroke=\"#000000\" points=\"222,-224.5 149,-224.5 149,-203.5 222,-203.5 222,-224.5\"/>\n<text text-anchor=\"middle\" x=\"185.5\" y=\"-210.9\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">TBackward</text>\n</g>\n<!-- 140303919175824&#45;&gt;140303919176080 -->\n<g id=\"edge5\" class=\"edge\">\n<title>140303919175824&#45;&gt;140303919176080</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M178.058,-203.3685C170.575,-192.6786 158.8774,-175.9678 149.7967,-162.9953\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"152.6141,-160.9167 144.0121,-154.7315 146.8794,-164.931 152.6141,-160.9167\"/>\n</g>\n<!-- 140303919177424 -->\n<g id=\"node7\" class=\"node\">\n<title>140303919177424</title>\n<polygon fill=\"#add8e6\" stroke=\"#000000\" points=\"234.5,-301 136.5,-301 136.5,-267 234.5,-267 234.5,-301\"/>\n<text text-anchor=\"middle\" x=\"185.5\" y=\"-287.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">encoder.0.weight</text>\n<text text-anchor=\"middle\" x=\"185.5\" y=\"-274.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\"> (50, 978)</text>\n</g>\n<!-- 140303919177424&#45;&gt;140303919175824 -->\n<g id=\"edge6\" class=\"edge\">\n<title>140303919177424&#45;&gt;140303919175824</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M185.5,-266.6966C185.5,-257.0634 185.5,-245.003 185.5,-234.8518\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"189.0001,-234.7912 185.5,-224.7913 182.0001,-234.7913 189.0001,-234.7912\"/>\n</g>\n<!-- 140303919523600 -->\n<g id=\"node8\" class=\"node\">\n<title>140303919523600</title>\n<polygon fill=\"#d3d3d3\" stroke=\"#000000\" points=\"285,-84.5 212,-84.5 212,-63.5 285,-63.5 285,-84.5\"/>\n<text text-anchor=\"middle\" x=\"248.5\" y=\"-70.9\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">TBackward</text>\n</g>\n<!-- 140303919523600&#45;&gt;140305286453520 -->\n<g id=\"edge7\" class=\"edge\">\n<title>140303919523600&#45;&gt;140305286453520</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M230.1793,-63.2281C212.9307,-53.0866 186.8914,-37.7764 167.3081,-26.2622\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"168.9445,-23.1642 158.5502,-21.1128 165.3966,-29.1985 168.9445,-23.1642\"/>\n</g>\n<!-- 140303919175248 -->\n<g id=\"node9\" class=\"node\">\n<title>140303919175248</title>\n<polygon fill=\"#add8e6\" stroke=\"#000000\" points=\"296,-161 207,-161 207,-127 296,-127 296,-161\"/>\n<text text-anchor=\"middle\" x=\"251.5\" y=\"-147.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\">decoder.weight</text>\n<text text-anchor=\"middle\" x=\"251.5\" y=\"-134.4\" font-family=\"Times,serif\" font-size=\"12.00\" fill=\"#000000\"> (978, 50)</text>\n</g>\n<!-- 140303919175248&#45;&gt;140303919523600 -->\n<g id=\"edge8\" class=\"edge\">\n<title>140303919175248&#45;&gt;140303919523600</title>\n<path fill=\"none\" stroke=\"#000000\" d=\"M250.7584,-126.6966C250.3456,-117.0634 249.8287,-105.003 249.3937,-94.8518\"/>\n<polygon fill=\"#000000\" stroke=\"#000000\" points=\"252.8875,-94.6322 248.9625,-84.7913 245.894,-94.932 252.8875,-94.6322\"/>\n</g>\n</g>\n</svg>\n"
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 89
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "PE3gQg0fCzp3"
      },
      "source": [
        "Now, we start training"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hvVjH47DwKWp"
      },
      "source": [
        "## training function\n",
        "def train_model(model, model_name , valid_point_epoch = 30 , model_save_epoch = 200 , iteration_n = 15000, print_train = True):\n",
        "    \n",
        "    criterion = nn.MSELoss()\n",
        "    optimizer=optim.Adam(list(model.parameters()), lr=learning_rate, weight_decay=0.0002)\n",
        "\n",
        "    start_time = time.time()\n",
        "    print('training start at %s '%(time.ctime(start_time)),'_'*50)\n",
        "    train_loss_record = []\n",
        "    valid_loss_record = []\n",
        "\n",
        "    for epoch in range(iteration_n): \n",
        "        inputs = next(train_loader)\n",
        "        if noise:\n",
        "            inputs_n = random_noise(inputs)\n",
        "        else:\n",
        "            inputs_n = inputs\n",
        "\n",
        "        inputs = torch.tensor(inputs).float()\n",
        "        inputs_n = torch.tensor(inputs_n).float()\n",
        "\n",
        "        # zero the parameter gradients\n",
        "        optimizer.zero_grad()\n",
        "        loss = criterion(model(inputs), inputs)\n",
        "        loss.backward()\n",
        "        optimizer.step()  # Update the parameters of the model\n",
        "\n",
        "        # print statistics\n",
        "        batch_loss = np.mean(loss.item())\n",
        "        #epochs.append(epoch+1)\n",
        "        #train_loss.append(batch_loss) \n",
        "        train_loss_record.append((epoch+1, batch_loss))\n",
        "\n",
        "        if print_train and epoch % 50 == 49:    # print every 50 epochs\n",
        "            print('epoch %d loss: %.3f' %\n",
        "                  (epoch + 1, np.mean(loss.item())))\n",
        "        if epoch % valid_point_epoch == valid_point_epoch-1:\n",
        "            valid_inputs = next(valid_loader)\n",
        "            valid_inputs = torch.tensor(valid_inputs).float()\n",
        "            valid_loss = criterion(model(valid_inputs), valid_inputs)\n",
        "            valid_loss_record.append((epoch+1, np.mean(valid_loss.item())))\n",
        "\n",
        "        # save check point models\n",
        "        if epoch % model_save_epoch == model_save_epoch-1:\n",
        "            save_models(work_dir,model,train_loss_record[-1][0],model_name)\n",
        "    print('_'*50,'\\nFinished Training at', time.ctime(time.time()), '\\ntotal training time: %.2f s' %(time.time()-start_time))\n",
        "\n",
        "    with open(f'{work_dir}/trained_model/{model_name}_train_loss_record.txt','w+') as ft:\n",
        "        ft.write('\\n'.join('%s\\t%s' % x for x in train_loss_record))\n",
        "    with open(f'{work_dir}/trained_model/{model_name}_valid_loss_record.txt','w+') as fv:\n",
        "        fv.write('\\n'.join('%s\\t%s' % x for x in valid_loss_record))\n",
        "    \n",
        "    %matplotlib inline\n",
        "    ## plot the training and validation curve\n",
        "    plt.close()\n",
        "    plt.plot(*zip(*train_loss_record),color='orange')\n",
        "    plt.plot(*zip(*valid_loss_record),color='green')\n",
        "    plt.xlabel(\"epoch\")\n",
        "    plt.ylabel(\"MSE\")\n",
        "    plt.savefig(f'{work_dir}/trained_model/{model_name}.pdf')\n",
        "    \n",
        "    return model"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 551
        },
        "id": "appZwE20tw5n",
        "outputId": "908a1d13-a950-45e8-f8b9-ab9846283ec5"
      },
      "source": [
        "## set hyper-parameters\n",
        "node_n=50\n",
        "layers = 1\n",
        "\n",
        "learning_rate = 0.002\n",
        "iteration_n = 5000\n",
        "\n",
        "model_name=f\"{layers}layers_{node_n}nodes_landmarkgene_DDRi_r\"\n",
        "valid_point_epoch = 50\n",
        "model_save_epoch = 500\n",
        "\n",
        "## instantiate the model\n",
        "DA = Adage_linear_decoder(train_data.shape[1], node_number=node_n, encoder_layers = layers)\n",
        "print(\"{}\\nADAGE Network Architecture:\\n{}\\n{}\\n\".format(\"=\"*31, \"~\"*31, DA))\n",
        "\n",
        "## train the model\n",
        "DA = train_model(DA, model_name, valid_point_epoch = valid_point_epoch, model_save_epoch = model_save_epoch, iteration_n = iteration_n, print_train = False)\n",
        "\n",
        "# calculate r2 to show relation between orininal gene expression and reconstruct gene expression\n",
        "data1=torch.tensor(valid_data).float()\n",
        "data2=DA(data1).detach().numpy()\n",
        "data1=data1.detach().numpy()\n",
        "\n",
        "import sklearn.metrics\n",
        "from sklearn.metrics import r2_score\n",
        "\n",
        "print(\"r_squre:\",r2_score(data1, data2))"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=50, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Tue Apr 13 20:20:16 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Tue Apr 13 20:20:33 2021 \n",
            "total training time: 16.96 s\n",
            "r_squre: 0.43279416700800943\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "fGyv6WoyTOMJ"
      },
      "source": [
        "# **2. Extracting high-weight genes**\n",
        "\n",
        "from the learned weights of each node, we extract genes corresponding to high weights, aka 2 or more stdev away from mean.\n",
        "\n",
        "$w_{gene} \\geq \\bar{w} \\pm 2 \\sigma $\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "EzFqBT66b9m7",
        "outputId": "b3e1830e-e191-4ee6-ded1-089717ceb249"
      },
      "source": [
        "# load the trained model\n",
        "node_n=50\n",
        "layers = 1\n",
        "\n",
        "learning_rate = 0.002\n",
        "epoch_n = 2500\n",
        "\n",
        "model_name_epoch=f\"{layers}layers_{node_n}nodes_landmarkgene_DDRi_r_epoch{epoch_n}\"\n",
        "DA = Adage_linear_decoder(train_data.shape[1], node_number=node_n, encoder_layers = layers)\n",
        "DA.load_state_dict(torch.load(f'{work_dir}/trained_model/{model_name_epoch}'))\n",
        "DA.eval()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "Adage_linear_decoder(\n",
              "  (encoder): Sequential(\n",
              "    (0): Linear(in_features=978, out_features=50, bias=True)\n",
              "    (1): ReLU(inplace=True)\n",
              "  )\n",
              "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
              ")"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 28
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "O4unuKK9B8u-"
      },
      "source": [
        "#print(\"weight matrix of the first layer:\",DA.encoder[0].weight.shape)\n",
        "#DA.encoder[0].weight"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 238
        },
        "id": "XQXMsU0Q81C9",
        "outputId": "2ce82b7b-2a77-48ca-fac1-572f64e34cc1"
      },
      "source": [
        "gene_info = pd.read_csv(gene_info_txt, sep = '\\t')\n",
        "print(gene_info.shape)\n",
        "#gene_info.head()\n",
        "landmark_gene = gene_info[gene_info['pr_is_lm'] == 1].astype(str)\n",
        "print(landmark_gene.shape)\n",
        "landmark_gene.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(12328, 5)\n",
            "(978, 5)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>pr_gene_id</th>\n",
              "      <th>pr_gene_symbol</th>\n",
              "      <th>pr_gene_title</th>\n",
              "      <th>pr_is_lm</th>\n",
              "      <th>pr_is_bing</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>780</td>\n",
              "      <td>DDR1</td>\n",
              "      <td>discoidin domain receptor tyrosine kinase 1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>7849</td>\n",
              "      <td>PAX8</td>\n",
              "      <td>paired box 8</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>25</th>\n",
              "      <td>6193</td>\n",
              "      <td>RPS5</td>\n",
              "      <td>ribosomal protein S5</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>43</th>\n",
              "      <td>23</td>\n",
              "      <td>ABCF1</td>\n",
              "      <td>ATP binding cassette subfamily F member 1</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>49</th>\n",
              "      <td>9552</td>\n",
              "      <td>SPAG7</td>\n",
              "      <td>sperm associated antigen 7</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "   pr_gene_id pr_gene_symbol  ... pr_is_lm pr_is_bing\n",
              "0         780           DDR1  ...        1          1\n",
              "1        7849           PAX8  ...        1          1\n",
              "25       6193           RPS5  ...        1          1\n",
              "43         23          ABCF1  ...        1          1\n",
              "49       9552          SPAG7  ...        1          1\n",
              "\n",
              "[5 rows x 5 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 29
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "z1cc6mPeAQHw",
        "outputId": "262f664a-7536-4258-b657-e96d68511033"
      },
      "source": [
        "landmark_gene['pr_gene_id'].to_list"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "<bound method IndexOpsMixin.tolist of 0          780\n",
              "1         7849\n",
              "25        6193\n",
              "43          23\n",
              "49        9552\n",
              "         ...  \n",
              "12184     5467\n",
              "12223     2767\n",
              "12224    23038\n",
              "12286    57048\n",
              "12321    79716\n",
              "Name: pr_gene_id, Length: 978, dtype: object>"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 21
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "Aq0iXGDKTWwb"
      },
      "source": [
        "def extract_decoder_high_weight_gene(DAmodel, dataframe = dataf, gene_info_file = gene_info_txt, stdev_cutoff = 2, save = True, \n",
        "                             savepath = f'{work_dir}/trained_model/{model_name_epoch}_high_weight_genes.txt'):\n",
        "    '''\n",
        "    This function read weight matrix, extract the corresponding gene id from input, and out put a summary dataframe\n",
        "    '''\n",
        "    ## load gene info\n",
        "    gene_info = pd.read_csv(gene_info_file, sep = '\\t')\n",
        "    gene_info['pr_gene_id'] = gene_info['pr_gene_id'].astype(str)\n",
        "\n",
        "    # get the gene id in the order of training data\n",
        "    gene_id = dataframe.columns.to_list()\n",
        "    # subset gene info and organize in the order of gene_id\n",
        "    gene_df = gene_info.set_index('pr_gene_id').loc[gene_id].reset_index()\n",
        "\n",
        "    #get the weitht matrix between genes and nodes\n",
        "    t_weight = DAmodel.decoder.weight.T\n",
        "\n",
        "    #calculate mean and std at each node of matrix\n",
        "    w_mean = t_weight.mean(dim=1).reshape(t_weight.shape[0],1)\n",
        "    w_std = t_weight.std(dim=1).reshape(t_weight.shape[0],1)\n",
        "\n",
        "    #check for the weights which is more than two std from the mean of each nodes\n",
        "    t_torf_HW=abs(t_weight-w_mean)- stdev_cutoff*w_std >= 0\n",
        "   \n",
        "    #extract and save the genes\n",
        "    dfs = []\n",
        "    for i in range(t_torf_HW.shape[0]):\n",
        "        node_hw = t_torf_HW[i].tolist()\n",
        "        if sum(node_hw) == 0:   # why some nodes give no true in node_hw???\n",
        "            print('node',i, '0 gene.')\n",
        "            continue\n",
        "        node_HWgene = gene_df[node_hw]\n",
        "        node_HWgene.loc[:,'gene_weight'] = t_weight[i][node_hw].tolist()\n",
        "        node_HWgene.loc[:,'node_id'] = i\n",
        "        dfs.append(node_HWgene)\n",
        "\n",
        "    HWgene = pd.concat(dfs)\n",
        "\n",
        "    if save:\n",
        "        HWgene.to_csv(savepath, sep='\\t',index=False)\n",
        "\n",
        "    return HWgene"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 527
        },
        "id": "N6aFMy89cNZh",
        "outputId": "ca152aa5-630d-4485-e4f5-0a4e29be7cbf"
      },
      "source": [
        "HWgene = extract_decoder_high_weight_gene(DA)\n",
        "print(HWgene.shape)\n",
        "\n",
        "HWgene.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "(2328, 7)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>pr_gene_id</th>\n",
              "      <th>pr_gene_symbol</th>\n",
              "      <th>pr_gene_title</th>\n",
              "      <th>pr_is_lm</th>\n",
              "      <th>pr_is_bing</th>\n",
              "      <th>gene_weight</th>\n",
              "      <th>node_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>52</th>\n",
              "      <td>3033</td>\n",
              "      <td>HADH</td>\n",
              "      <td>hydroxyacyl-CoA dehydrogenase</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0.090999</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>62</th>\n",
              "      <td>5211</td>\n",
              "      <td>PFKL</td>\n",
              "      <td>phosphofructokinase, liver</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0.087978</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>87</th>\n",
              "      <td>10730</td>\n",
              "      <td>YME1L1</td>\n",
              "      <td>YME1 like 1 ATPase</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>-0.094932</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>94</th>\n",
              "      <td>26020</td>\n",
              "      <td>LRP10</td>\n",
              "      <td>LDL receptor related protein 10</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>-0.119141</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>106</th>\n",
              "      <td>10146</td>\n",
              "      <td>G3BP1</td>\n",
              "      <td>GTPase activating protein (SH3 domain) binding...</td>\n",
              "      <td>1</td>\n",
              "      <td>1</td>\n",
              "      <td>0.100484</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "    pr_gene_id pr_gene_symbol  ... gene_weight  node_id\n",
              "52        3033           HADH  ...    0.090999        0\n",
              "62        5211           PFKL  ...    0.087978        0\n",
              "87       10730         YME1L1  ...   -0.094932        0\n",
              "94       26020          LRP10  ...   -0.119141        0\n",
              "106      10146          G3BP1  ...    0.100484        0\n",
              "\n",
              "[5 rows x 7 columns]"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 60
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "jTMU3t7X-ya8"
      },
      "source": [
        "# 3. Evaluate the learning results\n"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "htLieG0B_hYU"
      },
      "source": [
        "### 3.1 Compare gene sets with protein interaction database (STRING database) \n",
        "\n",
        "download and prepare the interaction database "
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "id": "YrDHj5L8_ljx",
        "outputId": "53946094-0772-4af3-da65-8817f72b6729"
      },
      "source": [
        "# download STRING database for human\n",
        "#stringdb_path = f'{work_dir}/9606.protein.links.full.v11.0.txt'\n",
        "stringdb_path = f'{work_dir}/9606.protein.links.v11.0.txt'\n",
        "if not os.path.isfile(stringdb_path):\n",
        "    #os.system(f'wget -O {stringdb_path}.gz https://stringdb-static.org/download/protein.links.full.v11.0/9606.protein.links.full.v11.0.txt.gz')\n",
        "    os.system(f'wget -O {stringdb_path}.gz https://stringdb-static.org/download/protein.links.v11.0/9606.protein.links.v11.0.txt.gz')\n",
        "    os.system(f'gunzip {stringdb_path}.gz')\n",
        "\n",
        "stringdb = pd.read_csv(stringdb_path, sep=' ') \n",
        "stringdb.head()\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>protein1</th>\n",
              "      <th>protein2</th>\n",
              "      <th>combined_score</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000272298</td>\n",
              "      <td>490</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000253401</td>\n",
              "      <td>198</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000401445</td>\n",
              "      <td>159</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000418915</td>\n",
              "      <td>606</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000327801</td>\n",
              "      <td>167</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "               protein1              protein2  combined_score\n",
              "0  9606.ENSP00000000233  9606.ENSP00000272298             490\n",
              "1  9606.ENSP00000000233  9606.ENSP00000253401             198\n",
              "2  9606.ENSP00000000233  9606.ENSP00000401445             159\n",
              "3  9606.ENSP00000000233  9606.ENSP00000418915             606\n",
              "4  9606.ENSP00000000233  9606.ENSP00000327801             167"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 23
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "korHibwoRTgw"
      },
      "source": [
        "get STRING IDs of the high weight genes\n",
        "\n",
        "(script adapted and modified from STRING API page: https://string-db.org/help/api/)\n",
        "\n"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "fkH4GKgR-xTm"
      },
      "source": [
        "import requests \n",
        "\n",
        "def symbol_to_ID(list_of_HUGO_symbol, string_api_url = \"https://string-db.org/api\", output_format = \"tsv-no-header\", method = \"get_string_ids\", species = 9606):  \n",
        "    '''A function to match HUGO gene symbol (used in L1000) to protein ID (used in STRING)\n",
        "    return a list of tuples (symbol, STRING_ID)\n",
        "    '''\n",
        "    \n",
        "    params = {\n",
        "        \"identifiers\" : \"\\r\".join(list_of_HUGO_symbol), # your protein list\n",
        "        \"species\" : species, # species NCBI identifier \n",
        "        \"limit\" : 1, # only one (best) identifier per input protein\n",
        "        \"echo_query\" : 1, # see your input identifiers in the output\n",
        "        \"caller_identity\" : \"ADAGE_project\" # your app name\n",
        "    }\n",
        "\n",
        "    ## Construct URL\n",
        "    request_url = \"/\".join([string_api_url, output_format, method])\n",
        "\n",
        "    ## Call STRING\n",
        "    results = requests.post(request_url, data=params)\n",
        "\n",
        "    ## Read and parse the results\n",
        "    IDmatch = []\n",
        "    for line in results.text.strip().split(\"\\n\"):\n",
        "        l = line.split(\"\\t\")\n",
        "        IDmatch.append((l[0], l[2]))\n",
        "        #input_identifier, string_identifier = l[0], l[2]\n",
        "        #print(\"Input:\", input_identifier, \"STRING:\", string_identifier, sep=\"\\t\")\n",
        "\n",
        "    return IDmatch\n",
        "\n",
        "def extract_STRINGinteractions(gene_df, save = False, save_path = f'{work_dir}/trained_model/stringdb_record.txt'):\n",
        "    '''the gene_df dataframe must have column \"node_id\" and \"STRING_ID\"'''\n",
        "    links = []\n",
        "    for i in gene_df['node_id'].unique():\n",
        "        s_ids = gene_df.loc[gene_df['node_id']==i]['STRING_ID'].tolist()\n",
        "\n",
        "        g_links = stringdb[stringdb['protein1'].isin(s_ids) & stringdb['protein2'].isin(s_ids)].copy()\n",
        "        #print(f'=== gene set #{i} ===\\n  {len(s_ids)} genes contain {len(g_links.index)} links')\n",
        "        if len(g_links.index) > 0:\n",
        "            g_links.loc[:,'node_id'] = i\n",
        "            links.append(g_links)\n",
        "        else:\n",
        "            print(f'=== Warning === gene set #{i}: {len(s_ids)} genes contain {len(g_links.index)} links')\n",
        "          \n",
        "    total_interactions = pd.concat(links)\n",
        "    if save:\n",
        "        total_interactions.to_csv(save_path,sep='\\t',index=False)\n",
        "\n",
        "    return total_interactions"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "j-l_nypHSf8N"
      },
      "source": [
        "# build a dictionary of gene symbol to StringID\n",
        "if os.path.isfile(f'{work_dir}/landmark_gene_STRING_ID.txt'):\n",
        "    STRINGID_dict = {}\n",
        "    with open(f'{work_dir}/landmark_gene_STRING_ID.txt') as f:\n",
        "        for line in f:\n",
        "            (key, val) = line[:-1].split('\\t')\n",
        "            STRINGID_dict[key] = val\n",
        "else:\n",
        "    landmark_gene_STRINGIDs = symbol_to_ID(landmark_gene_list['pr_gene_symbol'])\n",
        "    with open(f'{work_dir}/landmark_gene_STRING_ID.txt','w+') as f:\n",
        "        f.write('\\n'.join('%s\\t%s' % x for x in landmark_gene_STRINGIDs))\n",
        "    STRINGID_dict = dict(landmark_gene_STRINGIDs)"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "OAV-oupESuGG"
      },
      "source": [
        "Look at how many interactions in each gene set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "kkl8UcnbhbVv",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 221
        },
        "outputId": "d8a244c5-9cbd-43dc-d268-aeec674d6b25"
      },
      "source": [
        "# get STRING IDs of the high weight genes \n",
        "HWgene['STRING_ID'] = [STRINGID_dict[s] for s in HWgene['pr_gene_symbol'].to_list()]\n",
        "# interactions in each gene set\n",
        "total_interactions = extract_STRINGinteractions(HWgene,  save=True, save_path= f'{work_dir}/trained_model/{model_name_epoch}_HWgene_stringdb_records.txt')\n",
        "print(total_interactions.shape)\n",
        "total_interactions.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "(12934, 4)\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>protein1</th>\n",
              "      <th>protein2</th>\n",
              "      <th>combined_score</th>\n",
              "      <th>node_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>145577</th>\n",
              "      <td>9606.ENSP00000196371</td>\n",
              "      <td>9606.ENSP00000356015</td>\n",
              "      <td>961</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>145616</th>\n",
              "      <td>9606.ENSP00000196371</td>\n",
              "      <td>9606.ENSP00000320043</td>\n",
              "      <td>511</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>145628</th>\n",
              "      <td>9606.ENSP00000196371</td>\n",
              "      <td>9606.ENSP00000227868</td>\n",
              "      <td>300</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>145689</th>\n",
              "      <td>9606.ENSP00000196371</td>\n",
              "      <td>9606.ENSP00000300093</td>\n",
              "      <td>298</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>145710</th>\n",
              "      <td>9606.ENSP00000196371</td>\n",
              "      <td>9606.ENSP00000265333</td>\n",
              "      <td>282</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "                    protein1              protein2  combined_score  node_id\n",
              "145577  9606.ENSP00000196371  9606.ENSP00000356015             961        0\n",
              "145616  9606.ENSP00000196371  9606.ENSP00000320043             511        0\n",
              "145628  9606.ENSP00000196371  9606.ENSP00000227868             300        0\n",
              "145689  9606.ENSP00000196371  9606.ENSP00000300093             298        0\n",
              "145710  9606.ENSP00000196371  9606.ENSP00000265333             282        0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 33
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "5-i2HCmnYMhu"
      },
      "source": [
        "To understand the interaction-pair numbers, we estimate interaction-pair number with random genes in L1000:\n",
        "\n",
        "we generate same number of genesets with same number of random landmark genes in each set"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "8LrTrBi5YLp4",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "81ecc6be-00d2-4b43-c74f-dfedf0b2ab67"
      },
      "source": [
        "import random\n",
        "L1000gene = landmark_gene['pr_gene_symbol'].to_list()\n",
        "\n",
        "random_gene_list = []\n",
        "for i in HWgene['node_id'].unique():\n",
        "    gene_number = HWgene.node_id.value_counts()[i]\n",
        "    random_gene = random.sample(L1000gene, gene_number)\n",
        "    random_gene_stringID = [STRINGID_dict[s] for s in random_gene]\n",
        "    random_df = pd.DataFrame([(x[0],x[1],i) for x in zip(random_gene,random_gene_stringID)], columns=['pr_gene_symbol','STRING_ID','node_id'])\n",
        "    random_gene_list.append(random_df)\n",
        "\n",
        "random_genes = pd.concat(random_gene_list)\n",
        "random_genes.to_csv(f'{work_dir}/trained_model/{model_name_epoch}_random_landmark_genes.txt',sep='\\t',index=False)\n",
        "random_genes.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>pr_gene_symbol</th>\n",
              "      <th>STRING_ID</th>\n",
              "      <th>node_id</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>KIF20A</td>\n",
              "      <td>9606.ENSP00000378356</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>PCMT1</td>\n",
              "      <td>9606.ENSP00000356354</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>APP</td>\n",
              "      <td>9606.ENSP00000284981</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>MEF2C</td>\n",
              "      <td>9606.ENSP00000340874</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>PHGDH</td>\n",
              "      <td>9606.ENSP00000358417</td>\n",
              "      <td>0</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "  pr_gene_symbol             STRING_ID  node_id\n",
              "0         KIF20A  9606.ENSP00000378356        0\n",
              "1          PCMT1  9606.ENSP00000356354        0\n",
              "2            APP  9606.ENSP00000284981        0\n",
              "3          MEF2C  9606.ENSP00000340874        0\n",
              "4          PHGDH  9606.ENSP00000358417        0"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 34
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "Y66b8XXzqQP6"
      },
      "source": [
        "Compare the interaction numbers between high-weight gene sets and random gene sets"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hjVCeL_4gI2R"
      },
      "source": [
        "# calculate interactions\n",
        "random_gene_interaction = extract_STRINGinteractions(random_genes,save=True,save_path=f'{work_dir}/trained_model/{model_name_epoch}_random_landmark_stringdb_records.txt')\n",
        "\n",
        "# load interactions \n",
        "random_gene_interaction = pd.read_csv(f'{work_dir}/trained_model/{model_name_epoch}_random_landmark_stringdb_records.txt',sep='\\t')\n",
        "total_interactions = pd.read_csv(f'{work_dir}/trained_model/{model_name_epoch}_HWgene_stringdb_records.txt',sep='\\t')"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 595
        },
        "id": "cVNuvwJlDO61",
        "outputId": "9859eff7-50f1-4d1a-ae53-9d4dc7771ea6"
      },
      "source": [
        "%matplotlib inline\n",
        "from scipy import stats\n",
        "\n",
        "# compare number of interactions\n",
        "rdata = random_gene_interaction.node_id.value_counts()\n",
        "hdata = total_interactions.node_id.value_counts()\n",
        "\n",
        "plt.close()\n",
        "plt.scatter([rdata[i] for i in range(50)], [hdata[i] for i in range(50)])\n",
        "plt.xlabel(\"random landmark gene set\")\n",
        "plt.ylabel(\"high weight gene set\")\n",
        "plt.show()\n",
        "\n",
        "# perform t-test\n",
        "t, p = stats.ttest_ind(rdata, hdata)\n",
        "print('\\n\\nt-test p =', p)\n",
        "\n",
        "plt.close()\n",
        "data = [rdata, hdata]\n",
        "fig, ax = plt.subplots()\n",
        "ax.set_title('number of STRING DB interactions within one gene set')\n",
        "ax.set_xticklabels(['random gene set','high-weight gene set'])\n",
        "ax.boxplot(data)\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        },
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "t-test p = 0.00048527543726957707\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 332
        },
        "id": "Aoo7WIgvWy7X",
        "outputId": "05e87141-0bfe-4b3a-872a-e53b1b7f9ce2"
      },
      "source": [
        "%matplotlib inline\n",
        "# compare combined_score of interactions\n",
        "#rdata = random_gene_interaction.combined_score   # all combined_score\n",
        "#hdata = total_interactions.combined_score  # all combined_score\n",
        "\n",
        "rdata = [random_gene_interaction.loc[random_gene_interaction.node_id == i].combined_score.mean() for i in range(50)]   # average combined_score for each node\n",
        "hdata = [total_interactions.loc[total_interactions.node_id == i].combined_score.mean() for i in range(50)]  # average combined_score for each node\n",
        "\n",
        "# perform t-test\n",
        "t, p = stats.ttest_ind(rdata, hdata)\n",
        "print('\\n\\nt-test p =', p)\n",
        "\n",
        "## plot the training and validation curve\n",
        "plt.close()\n",
        "\n",
        "data = [rdata, hdata]\n",
        "fig, ax = plt.subplots()\n",
        "ax.set_title('STRING DB score of interactions')\n",
        "ax.set_xticklabels(['random gene set','high-weight gene set'])\n",
        "ax.boxplot(data)\n",
        "\n",
        "plt.show()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "\n",
            "\n",
            "t-test p = 0.9861972641297174\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "kpNSGHAJ_0oB"
      },
      "source": [
        "### 3.2 GO term analysis on gene sets with g:profile\n",
        "\n",
        "see documentation here: https://pypi.org/project/gprofiler-official"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "adcDVdPz_7r2",
        "outputId": "e1a9927b-1a62-4100-e41a-e20a532ee407"
      },
      "source": [
        "!pip install gprofiler-official"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "Collecting gprofiler-official\n",
            "  Downloading https://files.pythonhosted.org/packages/df/1b/5a87c1a1da8f601c00a0ce4dedb5aab8a5cad6a0f4a5062c4da22a045072/gprofiler_official-1.0.0-py3-none-any.whl\n",
            "Requirement already satisfied: requests in /usr/local/lib/python3.7/dist-packages (from gprofiler-official) (2.23.0)\n",
            "Requirement already satisfied: certifi>=2017.4.17 in /usr/local/lib/python3.7/dist-packages (from requests->gprofiler-official) (2020.12.5)\n",
            "Requirement already satisfied: idna<3,>=2.5 in /usr/local/lib/python3.7/dist-packages (from requests->gprofiler-official) (2.10)\n",
            "Requirement already satisfied: urllib3!=1.25.0,!=1.25.1,<1.26,>=1.21.1 in /usr/local/lib/python3.7/dist-packages (from requests->gprofiler-official) (1.24.3)\n",
            "Requirement already satisfied: chardet<4,>=3.0.2 in /usr/local/lib/python3.7/dist-packages (from requests->gprofiler-official) (3.0.4)\n",
            "Installing collected packages: gprofiler-official\n",
            "Successfully installed gprofiler-official-1.0.0\n"
          ],
          "name": "stdout"
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "id": "XFAZKUQ6I8jo",
        "outputId": "dcb12bbd-59e8-49cb-8fd0-a2a54fc7028d"
      },
      "source": [
        "from gprofiler import GProfiler\n",
        "#HWgene = pd.read_csv(f'{work_dir}/trained_model/onelayer_encoder_A549landmark_gene_epoch1000_high_weight_genes.txt',sep='\\t')\n",
        "gp = GProfiler(return_dataframe=True)\n",
        "terms = []\n",
        "for i in HWgene['node_id'].unique():\n",
        "    node_term = gp.profile(organism='hsapiens',query=HWgene.loc[HWgene['node_id']==i]['pr_gene_symbol'].tolist())\n",
        "    node_term.loc[:,'node_id'] = i\n",
        "    terms.append(node_term)\n",
        "HWterms = pd.concat(terms)\n",
        "HWterms.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1937, 15)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 99
        }
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "3D9d3lF7JrG0",
        "colab": {
          "base_uri": "https://localhost:8080/"
        },
        "outputId": "7cecdfb6-92de-4aa4-cb86-40cda0f00521"
      },
      "source": [
        "from gprofiler import GProfiler\n",
        "\n",
        "gp = GProfiler(return_dataframe=True)\n",
        "terms = []\n",
        "for i in random_genes['node_id'].unique():\n",
        "    node_term = gp.profile(organism='hsapiens',query=random_genes.loc[random_genes['node_id']==i]['pr_gene_symbol'].tolist())\n",
        "    node_term.loc[:,'node_id'] = i\n",
        "    terms.append(node_term)\n",
        "Rterms = pd.concat(terms)\n",
        "Rterms.shape"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/plain": [
              "(1331, 15)"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 100
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "w6qjeXsMEptf"
      },
      "source": [
        "# Appendix: automatic tools"
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "-famp-p5EuWH"
      },
      "source": [
        "trainings with different hyper parameters"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "hi1FOlw6Jtg7",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "ec5f635e-79f0-4907-addf-a716c8bb0c02"
      },
      "source": [
        "##### auto train with different hyper-parameters\n",
        "import sklearn.metrics\n",
        "from sklearn.metrics import r2_score\n",
        "\n",
        "iteration_n = 5000\n",
        "node_n=50\n",
        "for i in range(3):\n",
        "    for layers in [1, 5, 10]:\n",
        "\n",
        "        model_name = f\"{layers}layers_{node_n}nodes_landmarkgene_DDRi_r_withnoise\"\n",
        "        valid_point_epoch = 50\n",
        "        model_save_epoch = 500\n",
        "        learning_rate = 0.002\n",
        "\n",
        "        ## instantiate the model\n",
        "        DA = Adage_linear_decoder(train_data.shape[1], node_number=node_n, encoder_layers = layers)\n",
        "        print(\"{}\\nADAGE Network Architecture:\\n{}\\n{}\\n\".format(\"=\"*31, \"~\"*31, DA))\n",
        "\n",
        "        ## train the model\n",
        "        DA = train_model(DA, model_name, valid_point_epoch = valid_point_epoch, model_save_epoch = model_save_epoch, iteration_n = iteration_n, print_train = False)\n",
        "        # calculate r2 to show relation between orininal gene expression and reconstruct gene expression\n",
        "        data1=torch.tensor(train_data).float()\n",
        "        data2=DA(data1).detach().numpy()\n",
        "        data1=data1.detach().numpy()\n",
        "        data3=torch.tensor(valid_data).float()\n",
        "        data4=DA(data3).detach().numpy()\n",
        "        data3=data3.detach().numpy()\n",
        "\n",
        "        print(\"train_data r_squre:\",r2_score(data1, data2), \"\\n\",\"valid_data r_squre:\", r2_score(data3,data4))\n",
        "\n",
        "\n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=50, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 01:15:18 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 01:15:35 2021 \n",
            "total training time: 17.28 s\n",
            "train_data r_squre: 0.5177633970978205 \n",
            " valid_data r_squre: 0.4264618542563635\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=792, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Linear(in_features=792, out_features=606, bias=True)\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): Linear(in_features=606, out_features=420, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=420, out_features=234, bias=True)\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Linear(in_features=234, out_features=50, bias=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 01:15:36 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 01:24:53 2021 \n",
            "total training time: 557.19 s\n",
            "train_data r_squre: 0.4546760348391882 \n",
            " valid_data r_squre: 0.3578327529130308\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=885, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Linear(in_features=885, out_features=792, bias=True)\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): Linear(in_features=792, out_features=699, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=699, out_features=606, bias=True)\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Linear(in_features=606, out_features=513, bias=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Linear(in_features=513, out_features=420, bias=True)\n",
            "    (11): ReLU(inplace=True)\n",
            "    (12): Linear(in_features=420, out_features=327, bias=True)\n",
            "    (13): ReLU(inplace=True)\n",
            "    (14): Linear(in_features=327, out_features=234, bias=True)\n",
            "    (15): ReLU(inplace=True)\n",
            "    (16): Linear(in_features=234, out_features=141, bias=True)\n",
            "    (17): ReLU(inplace=True)\n",
            "    (18): Linear(in_features=141, out_features=50, bias=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 01:24:56 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 01:42:25 2021 \n",
            "total training time: 1048.89 s\n",
            "train_data r_squre: 0.3424585424748665 \n",
            " valid_data r_squre: 0.2542706266149375\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=50, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 01:42:30 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 01:42:47 2021 \n",
            "total training time: 17.12 s\n",
            "train_data r_squre: 0.5119593166151435 \n",
            " valid_data r_squre: 0.42183477016797316\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=792, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Linear(in_features=792, out_features=606, bias=True)\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): Linear(in_features=606, out_features=420, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=420, out_features=234, bias=True)\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Linear(in_features=234, out_features=50, bias=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 01:42:48 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 01:52:16 2021 \n",
            "total training time: 568.98 s\n",
            "train_data r_squre: 0.4604273118785725 \n",
            " valid_data r_squre: 0.369303655731898\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=885, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Linear(in_features=885, out_features=792, bias=True)\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): Linear(in_features=792, out_features=699, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=699, out_features=606, bias=True)\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Linear(in_features=606, out_features=513, bias=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Linear(in_features=513, out_features=420, bias=True)\n",
            "    (11): ReLU(inplace=True)\n",
            "    (12): Linear(in_features=420, out_features=327, bias=True)\n",
            "    (13): ReLU(inplace=True)\n",
            "    (14): Linear(in_features=327, out_features=234, bias=True)\n",
            "    (15): ReLU(inplace=True)\n",
            "    (16): Linear(in_features=234, out_features=141, bias=True)\n",
            "    (17): ReLU(inplace=True)\n",
            "    (18): Linear(in_features=141, out_features=50, bias=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 01:52:19 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 02:14:07 2021 \n",
            "total training time: 1307.96 s\n",
            "train_data r_squre: 0.3012142503873986 \n",
            " valid_data r_squre: 0.23177436111391572\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=50, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 02:14:13 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 02:14:30 2021 \n",
            "total training time: 16.79 s\n",
            "train_data r_squre: 0.5136921623712387 \n",
            " valid_data r_squre: 0.4243348128174115\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=792, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Linear(in_features=792, out_features=606, bias=True)\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): Linear(in_features=606, out_features=420, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=420, out_features=234, bias=True)\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Linear(in_features=234, out_features=50, bias=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 02:14:30 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 02:23:18 2021 \n",
            "total training time: 527.81 s\n",
            "train_data r_squre: 0.4671527982069239 \n",
            " valid_data r_squre: 0.36825746337688897\n",
            "===============================\n",
            "ADAGE Network Architecture:\n",
            "~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~\n",
            "Adage_linear_decoder(\n",
            "  (encoder): Sequential(\n",
            "    (0): Linear(in_features=978, out_features=885, bias=True)\n",
            "    (1): ReLU(inplace=True)\n",
            "    (2): Linear(in_features=885, out_features=792, bias=True)\n",
            "    (3): ReLU(inplace=True)\n",
            "    (4): Linear(in_features=792, out_features=699, bias=True)\n",
            "    (5): ReLU(inplace=True)\n",
            "    (6): Linear(in_features=699, out_features=606, bias=True)\n",
            "    (7): ReLU(inplace=True)\n",
            "    (8): Linear(in_features=606, out_features=513, bias=True)\n",
            "    (9): ReLU(inplace=True)\n",
            "    (10): Linear(in_features=513, out_features=420, bias=True)\n",
            "    (11): ReLU(inplace=True)\n",
            "    (12): Linear(in_features=420, out_features=327, bias=True)\n",
            "    (13): ReLU(inplace=True)\n",
            "    (14): Linear(in_features=327, out_features=234, bias=True)\n",
            "    (15): ReLU(inplace=True)\n",
            "    (16): Linear(in_features=234, out_features=141, bias=True)\n",
            "    (17): ReLU(inplace=True)\n",
            "    (18): Linear(in_features=141, out_features=50, bias=True)\n",
            "    (19): ReLU(inplace=True)\n",
            "  )\n",
            "  (decoder): Linear(in_features=50, out_features=978, bias=True)\n",
            ")\n",
            "\n",
            "training start at Wed Apr 14 02:23:21 2021  __________________________________________________\n",
            "__________________________________________________ \n",
            "Finished Training at Wed Apr 14 02:45:43 2021 \n",
            "total training time: 1342.36 s\n",
            "train_data r_squre: 0.3127739028582719 \n",
            " valid_data r_squre: 0.25601476444361976\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "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\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "G3BfAQNyFQY6"
      },
      "source": [
        "## Auto run for evaluating STRING interactions within gene sets:\n",
        "\n",
        "1. function to compare STRING interactions with random gene interactions"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "JFhXQ-mEFUF9"
      },
      "source": [
        "from scipy import stats\n",
        "def STRINGdb_compare(HWgene, random_genes, work_dir, model_name_epoch):\n",
        "    HWgene['STRING_ID'] = [STRINGID_dict[s] for s in HWgene['pr_gene_symbol'].to_list()]\n",
        "    total_interactions = extract_STRINGinteractions(HWgene,  save=True, save_path= f'{work_dir}/trained_model/{model_name_epoch}_HWgene_stringdb_records.txt')\n",
        "    random_gene_interaction = extract_STRINGinteractions(random_genes,save=True,save_path=f'{work_dir}/trained_model/{model_name_epoch}_random_landmark_stringdb_records.txt')\n",
        "\n",
        "    ## plot the number of interactions \n",
        "    rdata = random_gene_interaction.node_id.value_counts()\n",
        "    hdata = total_interactions.node_id.value_counts()\n",
        "    # perform t-test\n",
        "    t, p = stats.ttest_ind(rdata, hdata)\n",
        "    #print('\\n\\nt-test p =', p)\n",
        "\n",
        "    plt.close()\n",
        "    data = [rdata, hdata]\n",
        "    fig, ax = plt.subplots()\n",
        "    ax.set_title(f'number of STRING DB interactions within one gene set\\nt-test p ={p}')\n",
        "    ax.set_xticklabels(['random gene set','high-weight gene set'])\n",
        "    ax.boxplot(data)\n",
        "    plt.savefig(f'{work_dir}/trained_model/{model_name_epoch}_stringdb_interaction_number.pdf')\n",
        "\n",
        "    ## plot combined_score of interactions\n",
        "\n",
        "    rdata2 = [random_gene_interaction.loc[random_gene_interaction.node_id == i].combined_score.mean() for i in random_gene_interaction.node_id.unique()]   # average combined_score for each node\n",
        "    hdata2 = [total_interactions.loc[total_interactions.node_id == i].combined_score.mean() for i in total_interactions.node_id.unique()]  # average combined_score for each node\n",
        "    # perform t-test\n",
        "    t2, p2 = stats.ttest_ind(rdata2, hdata2)\n",
        "    \n",
        "    plt.close()\n",
        "    data2 = [rdata2, hdata2]\n",
        "    fig, ax = plt.subplots()\n",
        "    ax.set_title(f'STRING DB score of interactions\\nt-test p ={p2}')\n",
        "    ax.set_xticklabels(['random gene set','high-weight gene set'])\n",
        "    ax.boxplot(data2)\n",
        "    plt.savefig(f'{work_dir}/trained_model/{model_name_epoch}_stringdb_interaction_score.pdf')\n"
      ],
      "execution_count": null,
      "outputs": []
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "7hty6FSuFq8i",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 204
        },
        "outputId": "26361fb1-769d-496d-8712-fbe5ad828d15"
      },
      "source": [
        "## load stringdb ID dictionary\n",
        "with open(f'{work_dir}/landmark_gene_STRING_ID.txt') as f:\n",
        "    for line in f:\n",
        "        (key, val) = line[:-1].split('\\t')\n",
        "        STRINGID_dict[key] = val\n",
        "## load stringdb interactions\n",
        "stringdb = pd.read_csv(f'{work_dir}/9606.protein.links.v11.0.txt', sep=' ') \n",
        "stringdb.head()"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "execute_result",
          "data": {
            "text/html": [
              "<div>\n",
              "<style scoped>\n",
              "    .dataframe tbody tr th:only-of-type {\n",
              "        vertical-align: middle;\n",
              "    }\n",
              "\n",
              "    .dataframe tbody tr th {\n",
              "        vertical-align: top;\n",
              "    }\n",
              "\n",
              "    .dataframe thead th {\n",
              "        text-align: right;\n",
              "    }\n",
              "</style>\n",
              "<table border=\"1\" class=\"dataframe\">\n",
              "  <thead>\n",
              "    <tr style=\"text-align: right;\">\n",
              "      <th></th>\n",
              "      <th>protein1</th>\n",
              "      <th>protein2</th>\n",
              "      <th>combined_score</th>\n",
              "    </tr>\n",
              "  </thead>\n",
              "  <tbody>\n",
              "    <tr>\n",
              "      <th>0</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000272298</td>\n",
              "      <td>490</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>1</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000253401</td>\n",
              "      <td>198</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>2</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000401445</td>\n",
              "      <td>159</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>3</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000418915</td>\n",
              "      <td>606</td>\n",
              "    </tr>\n",
              "    <tr>\n",
              "      <th>4</th>\n",
              "      <td>9606.ENSP00000000233</td>\n",
              "      <td>9606.ENSP00000327801</td>\n",
              "      <td>167</td>\n",
              "    </tr>\n",
              "  </tbody>\n",
              "</table>\n",
              "</div>"
            ],
            "text/plain": [
              "               protein1              protein2  combined_score\n",
              "0  9606.ENSP00000000233  9606.ENSP00000272298             490\n",
              "1  9606.ENSP00000000233  9606.ENSP00000253401             198\n",
              "2  9606.ENSP00000000233  9606.ENSP00000401445             159\n",
              "3  9606.ENSP00000000233  9606.ENSP00000418915             606\n",
              "4  9606.ENSP00000000233  9606.ENSP00000327801             167"
            ]
          },
          "metadata": {
            "tags": []
          },
          "execution_count": 41
        }
      ]
    },
    {
      "cell_type": "markdown",
      "metadata": {
        "id": "bBg366MaFusz"
      },
      "source": [
        "2. loop to run different comparisons"
      ]
    },
    {
      "cell_type": "code",
      "metadata": {
        "id": "zjwETtmhGVZP",
        "colab": {
          "base_uri": "https://localhost:8080/",
          "height": 1000
        },
        "outputId": "1e1a2f2b-46b7-4fac-d848-cceab25f4473"
      },
      "source": [
        "## evaluate interactions in each gene set\n",
        "# load trained model\n",
        "epoch_n = 5000\n",
        "\n",
        "for node_n in [50,100]:\n",
        "    for layers in [1, 5, 10]:\n",
        "\n",
        "        model_name_epoch=f\"{layers}layers_{node_n}nodes_landmarkgene_DDRi_r_epoch{epoch_n}\"\n",
        "        print('===',model_name_epoch)\n",
        "        DA = Adage_linear_decoder(train_data.shape[1], node_number=node_n, encoder_layers = layers)\n",
        "        DA.load_state_dict(torch.load(f'{work_dir}/trained_model/{model_name_epoch}'))\n",
        "\n",
        "        # extract high weight gene\n",
        "        HWgene = extract_decoder_high_weight_gene(DA)\n",
        "        HWgene['STRING_ID'] = [STRINGID_dict[s] for s in HWgene['pr_gene_symbol'].to_list()]\n",
        "\n",
        "        # matplotlib histogram\n",
        "        t_weight = DA.decoder.weight.T\n",
        "        node_num = t_weight.shape[0]\n",
        "\n",
        "        w_mean = t_weight.mean(dim=1).reshape(t_weight.shape[0],1)\n",
        "        w_std = t_weight.std(dim=1).reshape(t_weight.shape[0],1)\n",
        "\n",
        "        plt.figure(figsize=(32,node_num))\n",
        "        for i in range(node_num):\n",
        "            hwgene_num = len(HWgene.loc[HWgene.node_id == i].index)\n",
        "\n",
        "            ax = plt.subplot(node_num/5, 5, i + 1)\n",
        "            ax.hist(t_weight[i].tolist(), color = '#56B4E9', edgecolor = 'white',bins = int(80))\n",
        "            ax.vlines([(w_mean-2*w_std)[i].item(),(w_mean+2*w_std)[i].item()], ymin=0, ymax=50, colors= 'red')\n",
        "            ax.vlines(w_mean[i].item(), ymin=0, ymax=50, colors= 'red',linestyles='dashed')\n",
        "            # Add labels\n",
        "            ax.set_title(f'weight distribution from node #{i}\\n number of high-weight genes: {hwgene_num}')\n",
        "            ax.set_xlabel('gene weight')\n",
        "            ax.set_ylabel('genes')\n",
        "\n",
        "        plt.tight_layout()\n",
        "        plt.savefig(f\"{work_dir}/trained_model/{layers}layers_{node_n}nodes_landmarkgene_DDRi_r_epoch{epoch_n}_weight_distribution.pdf\")\n",
        "        plt.close()\n",
        "\n",
        "\n",
        "\n",
        "        # generate random gene set\n",
        "        import random\n",
        "        L1000gene = landmark_gene_list['pr_gene_symbol'].to_list()\n",
        "        random_gene_list = []\n",
        "        for i in HWgene['node_id'].unique():\n",
        "            gene_number = HWgene.node_id.value_counts()[i]\n",
        "            random_gene = random.sample(L1000gene, gene_number)\n",
        "            random_gene_stringID = [STRINGID_dict[s] for s in random_gene]\n",
        "            random_df = pd.DataFrame([(x[0],x[1],i) for x in zip(random_gene,random_gene_stringID)], columns=['pr_gene_symbol','STRING_ID','node_id'])\n",
        "            random_gene_list.append(random_df)\n",
        "        random_genes = pd.concat(random_gene_list)\n",
        "        random_genes.to_csv(f'{work_dir}/trained_model/{model_name_epoch}_random_landmark_genes.txt',sep='\\t',index=False)\n",
        "        random_genes.head()\n",
        "\n",
        "        # compare the interactions\n",
        "        STRINGdb_compare(HWgene, random_genes, work_dir, model_name_epoch)\n",
        " \n"
      ],
      "execution_count": null,
      "outputs": [
        {
          "output_type": "stream",
          "text": [
            "=== 1layers_50nodes_landmarkgene_DDRi_r_epoch5000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "=== 5layers_50nodes_landmarkgene_DDRi_r_epoch5000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "=== Warning === gene set #10: 1 genes contain 0 links\n",
            "=== Warning === gene set #13: 1 genes contain 0 links\n",
            "=== Warning === gene set #22: 1 genes contain 0 links\n",
            "=== Warning === gene set #26: 2 genes contain 0 links\n",
            "=== Warning === gene set #28: 1 genes contain 0 links\n",
            "=== Warning === gene set #31: 1 genes contain 0 links\n",
            "=== Warning === gene set #46: 1 genes contain 0 links\n",
            "=== Warning === gene set #10: 1 genes contain 0 links\n",
            "=== Warning === gene set #13: 1 genes contain 0 links\n",
            "=== Warning === gene set #17: 3 genes contain 0 links\n",
            "=== Warning === gene set #22: 1 genes contain 0 links\n",
            "=== Warning === gene set #26: 2 genes contain 0 links\n",
            "=== Warning === gene set #28: 1 genes contain 0 links\n",
            "=== Warning === gene set #31: 1 genes contain 0 links\n",
            "=== Warning === gene set #46: 1 genes contain 0 links\n",
            "=== Warning === gene set #47: 3 genes contain 0 links\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "=== 10layers_50nodes_landmarkgene_DDRi_r_epoch5000\n",
            "node 2 0 gene.\n",
            "node 23 0 gene.\n",
            "node 29 0 gene.\n",
            "node 36 0 gene.\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "node 39 0 gene.\n",
            "node 40 0 gene.\n",
            "node 41 0 gene.\n",
            "=== Warning === gene set #21: 7 genes contain 0 links\n",
            "=== Warning === gene set #22: 1 genes contain 0 links\n",
            "=== Warning === gene set #21: 7 genes contain 0 links\n",
            "=== Warning === gene set #22: 1 genes contain 0 links\n",
            "=== 1layers_100nodes_landmarkgene_DDRi_r_epoch5000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "=== 5layers_100nodes_landmarkgene_DDRi_r_epoch5000\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "=== Warning === gene set #8: 1 genes contain 0 links\n",
            "=== Warning === gene set #17: 1 genes contain 0 links\n",
            "=== Warning === gene set #25: 1 genes contain 0 links\n",
            "=== Warning === gene set #45: 1 genes contain 0 links\n",
            "=== Warning === gene set #47: 1 genes contain 0 links\n",
            "=== Warning === gene set #48: 1 genes contain 0 links\n",
            "=== Warning === gene set #56: 1 genes contain 0 links\n",
            "=== Warning === gene set #59: 2 genes contain 0 links\n",
            "=== Warning === gene set #76: 1 genes contain 0 links\n",
            "=== Warning === gene set #77: 1 genes contain 0 links\n",
            "=== Warning === gene set #87: 1 genes contain 0 links\n",
            "=== Warning === gene set #94: 2 genes contain 0 links\n",
            "=== Warning === gene set #3: 3 genes contain 0 links\n",
            "=== Warning === gene set #8: 1 genes contain 0 links\n",
            "=== Warning === gene set #17: 1 genes contain 0 links\n",
            "=== Warning === gene set #25: 1 genes contain 0 links\n",
            "=== Warning === gene set #45: 1 genes contain 0 links\n",
            "=== Warning === gene set #47: 1 genes contain 0 links\n",
            "=== Warning === gene set #48: 1 genes contain 0 links\n",
            "=== Warning === gene set #50: 4 genes contain 0 links\n",
            "=== Warning === gene set #56: 1 genes contain 0 links\n",
            "=== Warning === gene set #59: 2 genes contain 0 links\n",
            "=== Warning === gene set #76: 1 genes contain 0 links\n",
            "=== Warning === gene set #77: 1 genes contain 0 links\n",
            "=== Warning === gene set #87: 1 genes contain 0 links\n",
            "=== Warning === gene set #94: 2 genes contain 0 links\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1596: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  self.obj[key] = _infer_fill_value(value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1743: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(ilocs[0], value)\n",
            "/usr/local/lib/python3.7/dist-packages/pandas/core/indexing.py:1763: SettingWithCopyWarning: \n",
            "A value is trying to be set on a copy of a slice from a DataFrame.\n",
            "Try using .loc[row_indexer,col_indexer] = value instead\n",
            "\n",
            "See the caveats in the documentation: https://pandas.pydata.org/pandas-docs/stable/user_guide/indexing.html#returning-a-view-versus-a-copy\n",
            "  isetter(loc, value)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "stream",
          "text": [
            "=== 10layers_100nodes_landmarkgene_DDRi_r_epoch5000\n",
            "node 59 0 gene.\n",
            "node 67 0 gene.\n",
            "node 74 0 gene.\n",
            "node 84 0 gene.\n",
            "node 95 0 gene.\n",
            "node 96 0 gene.\n",
            "=== Warning === gene set #2: 1 genes contain 0 links\n",
            "=== Warning === gene set #3: 1 genes contain 0 links\n",
            "=== Warning === gene set #11: 2 genes contain 0 links\n",
            "=== Warning === gene set #21: 2 genes contain 0 links\n",
            "=== Warning === gene set #32: 1 genes contain 0 links\n",
            "=== Warning === gene set #33: 2 genes contain 0 links\n",
            "=== Warning === gene set #51: 1 genes contain 0 links\n",
            "=== Warning === gene set #54: 2 genes contain 0 links\n",
            "=== Warning === gene set #61: 1 genes contain 0 links\n",
            "=== Warning === gene set #76: 1 genes contain 0 links\n",
            "=== Warning === gene set #90: 2 genes contain 0 links\n",
            "=== Warning === gene set #92: 1 genes contain 0 links\n",
            "=== Warning === gene set #97: 2 genes contain 0 links\n",
            "=== Warning === gene set #2: 1 genes contain 0 links\n",
            "=== Warning === gene set #3: 1 genes contain 0 links\n",
            "=== Warning === gene set #13: 2 genes contain 0 links\n",
            "=== Warning === gene set #17: 5 genes contain 0 links\n",
            "=== Warning === gene set #22: 6 genes contain 0 links\n",
            "=== Warning === gene set #25: 3 genes contain 0 links\n",
            "=== Warning === gene set #32: 1 genes contain 0 links\n",
            "=== Warning === gene set #33: 2 genes contain 0 links\n",
            "=== Warning === gene set #48: 3 genes contain 0 links\n",
            "=== Warning === gene set #49: 5 genes contain 0 links\n",
            "=== Warning === gene set #51: 1 genes contain 0 links\n",
            "=== Warning === gene set #54: 2 genes contain 0 links\n",
            "=== Warning === gene set #61: 1 genes contain 0 links\n",
            "=== Warning === gene set #64: 5 genes contain 0 links\n",
            "=== Warning === gene set #72: 5 genes contain 0 links\n",
            "=== Warning === gene set #76: 1 genes contain 0 links\n",
            "=== Warning === gene set #90: 2 genes contain 0 links\n",
            "=== Warning === gene set #92: 1 genes contain 0 links\n",
            "=== Warning === gene set #97: 2 genes contain 0 links\n"
          ],
          "name": "stdout"
        },
        {
          "output_type": "stream",
          "text": [
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n",
            "/usr/local/lib/python3.7/dist-packages/numpy/core/_asarray.py:83: VisibleDeprecationWarning: Creating an ndarray from ragged nested sequences (which is a list-or-tuple of lists-or-tuples-or ndarrays with different lengths or shapes) is deprecated. If you meant to do this, you must specify 'dtype=object' when creating the ndarray\n",
            "  return array(a, dtype, copy=False, order=order)\n"
          ],
          "name": "stderr"
        },
        {
          "output_type": "display_data",
          "data": {
            "image/png": "iVBORw0KGgoAAAANSUhEUgAAAXcAAAEXCAYAAABWNASkAAAABHNCSVQICAgIfAhkiAAAAAlwSFlzAAALEgAACxIB0t1+/AAAADh0RVh0U29mdHdhcmUAbWF0cGxvdGxpYiB2ZXJzaW9uMy4yLjIsIGh0dHA6Ly9tYXRwbG90bGliLm9yZy+WH4yJAAAgAElEQVR4nO3df7xVVZ3/8ddbRFDUEL05Aipmlgg1WGhWVGKlYU7a5M/MtJjImaJxtAblzrd0Cqcfjma3jERMTUEdpybHoizBFCc18CeKpqUmiIoKKBgI+Pn+sdbBzfHce8/9xbls3s/HYz/uPmutvfc65+z7OXuvvfbaigjMzKxctmp0BczMrPs5uJuZlZCDu5lZCTm4m5mVkIO7mVkJObibmZWQg7vZJqDkx5KWSbqzRv6Jkm5sRN26StJkSZc0uh62MQf3zYCkMZL+T9IKSS9Iuk3SAfmfamWeVktaX3j9QF42JK3KaYslnS+pT2HdN0v6hzx/cC5/UdX250o6pfB6N0nTJD2V1/tnSZdJ2reV+h8s6dVC3RZJulbSAVXlinV9TtJMSQO78aNspDHAh4GhEXFgdWZEXBURh9azIkmnSJrb3RWsc9sHS1pUTIuIcyPiHxpRH2udg3svJ2lH4AagBRgEDAHOAdbkf6rtI2J74FTg95XXETGisJq/zWU+ABwHfLaNTa4CTpI0rJX67Az8H7Ad8D5gB+AdwO9Iwas1T+U67AAcBDwE3Crpg1XlKnV9E7ATcHYb6+wx+Ui7O/8/9gQej4hV3bjOTpG0daPrYD3Pwb33ewtARMyMiPUR8deIuDEi7uvoiiLiUeA2YFQbxZYDlwFfayX/X4AXgZMi4k+RLI+IH0dESx11iIhYFBFfBS4BvtVKuReB64H9WluXpEn5bOQlSQ9Xfigk9clnNX/KefMl7Z7z3iPpD/ks6A+S3lNY382Spki6DXgZeJOkfSX9Jp8xPSzp2DbqM1jS9bnso5I+l9PH5/f67nxWck6NZTc6Gs9nMadKekTSckk/yD84w4GphXUtz+X7STpP0l8kPSNpqqRtc97B+WxpkqSngR9L2knSDZKW5qaiGyQNLWx/kFIz0lM5/38kDQBmAYMLZ2GDJZ0t6crCsh+T9ECu9825zpW8xyV9WdJ9+Tu4RlL/nLdLrsfy/Bne2s0/sFsUf3C93x+B9ZIulzRO0k6dXVFuNnkf8Gg7RacAn5D01hp5HwJ+FhGvdrYeBT8F3pGDxkby+zwKuL3WgrluXwQOiIgdgMOAx3P26cAJwOHAjqQzlZclDQJ+AXwP2Bk4H/hFPhupOAmYQDrDWAr8BpgBvBE4HrhIUms/OFcDi4DBwNHAuZIOiYjpbHxm1doPZ7UjgAOAtwPHAodFxMKqdVWarb5JOhAYBbyZdIb31cK6/oZ05rdnfn9bAT/Or/cA/gp8v1D+J6SzsxH5vV+QzzrGkc/C8vRUscKS3gLMBE4DmoBfAv8raZtCsWOBjwB75fd2Sk4/I39+TcCuwGTA46N0koN7L5ePYMeQdvJpwNJ8dLhrB1Zzl6RVwELgZuCitgpHxNOko8N/r5G9C/B05UU+Sluej5A7ekHwKUBAsV39rnw0+hwp6PyolWXXA/2A/ST1jYjHI+JPOe8fgH+LiIfzmcK9EfE88FHgkYj4SUSsi4iZpOahvyus97KIeCAi1pEC0OP5rGRdRNwN/DdwTHVl8pnBe4FJEbE6Iu4hHa1/uoOfSdE381nRX4A5tHLGJUmkgP0vEfFCRLwEnEv6Map4FfhaRKzJZ3/PR8R/R8TLufwUUrMdknYjBfFTI2JZRKyNiN/VWefjgF9ExG8iYi1wHrAt8J5Cme9FxFMR8QLwv4X3tRbYDdgzb/PW8OBXnebgvhmIiIURcUpEDAVGko4Mv9uBVbwD2J70j/cu4HVHyjV8CzhM0t9WpT9P+ges1O36fPT4L8A2dMwQ0o/W8mJd8/r6Az8ktcv3r14wNzGdRmqTf1bS1ZIG5+zdgT9VL0P63J6oSnsi16PiycL8nsC78o/X8vyjcyLpKLjWuiuBtbV1d9TThfmXSd9hLU2ko+z5hXr+KqdXLI2I1ZUXkraT9CNJT0h6EbgFGKh0sX33/F6WdaLOG33G+QzvSTb+HFp7X98hnVXeqHSR/sxObN8yB/fNTEQ8RGoTH9nB5SIirgV+z8an662Vf570A/L1qqybgKO6qS3048BdtS4y5qO+S0in7jXfa0TMiIgxpCAcvNZ+/ySwd41Fnspli/YAFhdXW5h/EvhdRAwsTNtHxD+2su5BknZoY93dpfpo9jlSs8qIQj3fkC9Mt7bMGcBbgXdFxI7A+3O6SO97kGr3VGrvSHqjzzifVexOHZ9DRLwUEWdExJuAjwGn6/UX3K1ODu69XL6gd0blYlc+/T+BVtqi6/BN4HOSah19VjufdDo9vCptJ+AnkvbOF/l2oO2LtBvk8kMkfY3UfDK5lXJ9gM+Qgtafa+S/VdIhkvoBq3O5ynWAS4CvS9onb+/tuV39l8BbJH1S0taSjiNdsL2hlerekMufJKlvng4oXiCsiIgnSb2I/kNSf0lvB8YDV1aX7QbPAEMr7dj56HgacIGkNwLkz/iwNtaxA+kzW56vRWy4DhARS0gXTi/KF177SqoE/2eAnSW9oZX1Xgt8VNIHJfUl/YisIX02bZJ0hKQ35x+EFaSmt+64trNFcnDv/V4iNaXckdvNbwcWkP5pOiwi7iedgn+ljrIvAt8mXYirpD1H6sq4Gpib63cPKVjUOqKtGCxpJbAS+APwNuDgiKhup783l1sGnAx8PLfNVutH+qF6jnSa/0bgrJx3PinI3Ejq2TMd2DafjRxB+uyeB/4VOCK/p1rv/yXgUFLb9VN5O9/K267lBGBYLvszUhv3b1sp2xWzgQeApyVV6j6J1KRxe25m+S3pyLw13yW1hT9H2qd+VZV/EqkN/CHgWVITWOXMcSbw59wENLi4UEQ8DHyK1HX3OdL1jL+LiFfqeF/75HqvJJ1hXhQRc+pYzmqQr1eYmZWPj9zNzErIwd3MrIQc3M3MSsjB3cyshBzczcxKyMG9l8oDLH2onTIbhuvt4rZeN4zrpiJplNLAXi/nv632l5d0paQlkl6U9Mfq9y7pWEkL81AID0o6qpAnSd9QGmhsRf7sRlQt/yFJdykNO7xIhUHC2qqnpIFKY/88m6ezq9b7Hkl35nrdJ2lMVf4n852iq5QG6BpUyFtZNa2X1FLI307SRUpDJK+QdEshb1bVsq9Iur+Q/3VJ90taV6POH1Ua6nm5pKclXaLCDVpKg5Q9kt/TQ5I+Xch7i6SfKw1K9oKkX6swTlF734Wkb0t6Mn/PT0iqeS+EtSMiPPXCiTQI1ofaKXMz8A/dsK2DgUUNeI/bkG5V/xdS3/Ev5dfbtFJ+BNAvz+9L6nf+zvx6CPAKaUwUkcaReRl4Y84/ltT//E1AH+A/SHfHVta9H6k/9zhga9LAYnvXU0/SAFz/RRoCYBhp6IPP5LxBpD71x+TtforUh3+nwnt6iXSH6PakQcqubuX9b0/qA/7+QtqVpAHLmvL639nO/vLVwuuT8/v9OXB2VdlPksbW2Y5009osYGoh/5z8HWxFug9jGfCenHcg6QauQUBf0l3ODxWWbe+7eCswoPC9PgD8faP/Jze3qeEV8FTjS0kj8r1KuoNwJfCvNcpMId3BtzqX+X5O35c0kuELwMPAsYVlDgcezMFkMfBl0jgzlbs7KzcZDa6xvctIg4n9Ji//O9IAT115n4fmeqiQ9hfgI3Us+1ZgSeX95QDzbFWZpcC78/wk4NpC3ghgdeH1DODrnakn6WadAwp5k4Fb8/wRwANV6/sjMD7PnwvMKOTtTfqR2qFGPU4m3a1buT9lX9JNWjvW8XkNy/vLsBp5V1IV3GuU+Xvg/jbyrwfOaCVvEGnYgp3r+S6qlh0C3F/rf8BT25ObZXqhiDiJFDz+LtJYJt+uUaYZuBX4Yi7zRaWhc9saonY68PlIQ+SOBGZHHcO4FpxIOgrbhXRX6lWtvQcVBtuqMVUGhBoB3Bf5vzi7L6e3tt6LJL1MunNyCWlIAYB5wEKlUSr75CaZNXl9kI5u985NBn1JgbJ4V+ZBef3356afKwvNI/XUU1XzI1vJq84fAdxbyYg0suUr5HH8q5wMXFGox4GkM4hzcrPM/ZI+UWM5SKNT3hoRj7eS3573k46gX0dp3PgDWsvPyz4d6Q5haP+7QNKZSncqLyIdgMzoZL23WA7u5XIEbQ9Ru5Y0RO6OkYZyvauD6/9FRNwSEWuAZtIDI3avVTA2HmyrevpmLrY9aQyRohWkoQxqioh/yvnvI40HvyanrweuIAWBNfnv5+O1QcmWkIZLeJh0pnIMqZmlYijplvtPkG6D35Z0C3099fwVcKakHSS9mTR+/HY57/ekoRdOUBqj5WTS0Xklv67PQNKepCF5L6+q88hcfjBpfPvLVWPsG1Jwv6xGerskfZgUgFsbcG4q6Qfq1zWWHQr8gDTGfkV73wV5H6k85esnvP4zsnY4uG8mlJ6sU7kw1toFpvaGqP0EqWnmCUm/k/TuDlZjw3C4EbGS1PQzuPXi7VpJephG0Y6kZp9WRXoi1VxScPtHSBdDSePgHExqI/8AcEnhwudXSUeXu5OGEz4HmC2pEmT/Cvw4Iv6Y39u5pM+qnnp+KS//CKn9eibpiJN8tHokKbg9Q2rH/m0lvwOfwUnA3Ih4rJD2V9IP9jci4pVIY67PITUjbZAv4P4NcB0dJOkg0g/l0RHxxxr53yH9wBxbdWaDpCbS+D4XRRo7v6K97wLYMJLp3fl9vu7pVdY2B/fea6N/lIg4tdBscm6tMrQzRG1E/CEijiQ12fwPaXCtWutpzYajdEnbk9pSazbh1OjlsbLGj9MDwNslFZst3k7rp/fVtua1oX1HAbdExLyIeDUi/gDcQXpyVCX/mkiP+FsXEZeRLhRWmqzuY+PPoTjfZj0jPSDjxIj4m0jPrt0KuHPDiiJ+FxEHRMQgUpDet5D/ALBhzHxJbyJdtK0OpJ9m46P2Sp2r1fouTwZ+mn+06iZpf1Jb+mcj4qYa+eeQmvQOjTTIXDFvJ1Jgvz4iplQt2t53Ua34PVu9Gt3o76n2RBqpb0I7Za4Gzi283oHUBnsSqZdCX9IR0nDS0eyJwBty2fHAE3l+X9LR0Rva2NZlpIt3Y/K6LgBu6+J7rPRC+WdSQPsirfSW4bVrCNuTelgcRnqY98dy/gdIFzZH5df7k3qpHJpff43UFLArKfielJcfmPM/CzxG6sGxHemH7yf11JMUeHbO9RqX6zGiUPf983exI2k0xtsKeSPy5/o+UtvylVT1liENu7yKqouseZ2PAv+PFADfSzri37dQZltSk8YhNT7TvqQj5xnAN/J8n5w3knSmcVwr391ZpDOVv6mRtyPpx+v7rSzb6neRX3+eFOxFuq6wBPhSo/8nN7ep4RXw1MoXk07l/0J6StGXWynzbtIR3jLSo8sg9SL5BamnyPOk4WFH5QD1q1z2RdKwu2MK67o0l19O+71lVpKGDd6rG97n/sB80o/LXcD+hbzJwKw830TqobM81/9+4HNV6/piDnYvkXqVnFHI609q+12Sl7+Lql45pFP/pXn6Cbm7Yh31rHTte5l0ofmwqvXOzAF2BXANuXtmIf+T+bteRWrWGVSV/yPyD02Nz28EqV1/Fakn1Mer8k8g/RCpxrKXkY70i9MpOe/HbNyDaiWFXj+57Jqq/Mk57+Scv6oqf4/2vgtScP8VqclvJWn/nlyr/p7anjzkr9VF0mWkvvD/1ui6mFn73OZuZlZCDu5mZiXkZhkzsxLykbuZWQlt3egKAOyyyy4xbNiwRlfDzGyzMn/+/OcioqlWXq8I7sOGDWPevHmNroaZ2WZF0hOt5dXVLKM0tvj9ku6RNC+nnZ3HY74nT4cXyp8l6VFJD0s6rOtvwczMOqIjR+5jI+K5qrQLIuK8YkIegfB40s0Vg4HfSnpLpIGdzMxsE+iJC6pHkm6fXhNpkKNHSbcQm5nZJlJvcA/gRqXHi00opH9R6bFhl+aBgiANrv9kocyinLYRSRMkzZM0b+nSpZ2qvJmZ1VZvcB8TEe8gDYr0BUnvB35IGjBpFGmMiP/syIYj4uKIGB0Ro5uaal7sNTOzTqoruEfE4vz3WeBnwIER8UykcbVfBabxWtPLYgpDw5LG3F7cfVU2s83NzJkzGTlyJH369GHkyJHMnDmz/YWsS9oN7pIGKD/1PD/G7VBggaTdCsU+DizI89cDx0vqJ2kv0lNt7sTMtkgzZ86kubmZlpYWVq9eTUtLC83NzQ7wPazd4QfywwN+ll9uTXqY7xRJPyE1yQTwOOmRZkvyMs2k8bHXAadFxKy2tjF69OhwP3ezcho5ciQtLS2MHTt2Q9qcOXOYOHEiCxYsaGNJa4+k+RExumZebxhbxsHdrLz69OnD6tWr6du374a0tWvX0r9/f9avdw/prmgruHtsGTPrUcOHD2fu3Lkbpc2dO5fhw2s9x9u6i4O7mfWo5uZmxo8fz5w5c1i7di1z5sxh/PjxNDc3N7pqpdYrxpYxs/I64YQTAJg4cSILFy5k+PDhTJkyZUO69Qy3uZuZbabc5m5mtoVxcDczKyEHdzOzEnJwNzMrIQd3M7MScnA3MyshB3czsxJycC8RD6tqZhW+Q7UkKsOqTp8+nTFjxjB37lzGjx8P4DsBzbZAvkO1JDysqtmWx0P+bgE8rKrZlsfDD2wBPKyqmRU5uJeEh1U1s6K6LqhKehx4CVgPrIuI0ZIGAdcAw0iP2Ts2IpZJEnAhcDjwMnBKRNzV/VW3Ig+ramZFdbW55+A+OiKeK6R9G3ghIr4p6Uxgp4iYJOlwYCIpuL8LuDAi3tXW+t3mbmbWcT3V5n4kcHmevxw4qpB+RSS3AwMl7daF7ZiZWQfVG9wDuFHSfEkTctquEbEkzz8N7JrnhwBPFpZdlNPMzGwTqfcmpjERsVjSG4HfSHqomBkRIalDfSrzj8QEgD322KMji5qZWTvqOnKPiMX577PAz4ADgWcqzS3577O5+GJg98LiQ3Na9TovjojRETG6qamp8+/AzHo9D42x6bUb3CUNkLRDZR44FFgAXA+cnIudDPw8z18PfFrJQcCKQvONmW1hKkNjtLS0sHr1alpaWmhubnaA72Ht9paR9CbS0TqkZpwZETFF0s7AtcAewBOkrpAv5K6Q3wc+QuoK+ZmIaLMrjHvLmJWXh8boOR5+wMwaxkNj9BwPP2BmDeOhMRrDwd3MepSHxmgMj+duZj3KQ2M0htvczcw2U25zNzPbwji4m1mPmzhxIv3790cS/fv3Z+LEiY2uUuk5uJtZj5o4cSJTp07l3HPPZdWqVZx77rlMnTrVAb6Huc3dzHpU//79Offcczn99NM3pJ1//vlMnjyZ1atXN7Bmmz/fxGRmDSOJVatWsd12221Ie/nllxkwYAC9If5sznxB1cwapl+/fkydOnWjtKlTp9KvX78G1WjL4H7uZtajPve5zzFp0iQATj31VKZOncqkSZM49dRTG1yzcnNwN7Me1dLSAsDkyZM544wz6NevH6eeeuqGdOsZbpYpEY+Zbb1VZbjfiNgw7K/1LB+5l0RlzOzp06czZswY5s6dy/jx4wF8m7fZFsi9ZUrCY2abbXncFXIL4DGzzbY87gq5BfCY2WZW5OBeEh4z28yK6r6gKqkPMA9YHBFHSLoM+ACwIhc5JSLuyc9QvRA4nPQM1VMi4q7urbZV85jZZlbUkSP3fwYWVqV9JSJG5emenDYO2CdPE4Afdr2aVo8TTjiBBQsWsH79ehYsWODAbr2GR4Xc9OoK7pKGAh8FLqmj+JHAFZHcDgyUtFsX6mhmmzGPCtkY9R65fxf4V+DVqvQpku6TdIGkykARQ4AnC2UW5bSNSJogaZ6keUuXLu1ovc1sMzFt2jSOO+44Lr30UnbYYQcuvfRSjjvuOKZNm9boqpVau8Fd0hHAsxExvyrrLGBf4ABgEDCpIxuOiIsjYnREjG5qaurIoma2GVmzZg233XbbhrtUW1pauO2221izZk2jq1Zq9Ry5vxf4mKTHgauBQyRdGRFLctPLGuDHwIG5/GJg98LyQ3OamW2BJDFu3DjGjh1L3759GTt2LOPGjSP1vbCe0m5wj4izImJoRAwDjgdmR8SnKu3ouXfMUUDlNsjrgU8rOQhYERFLeqb6ZtbbRQTTpk3j/PPP5+WXX+b8889n2rRpHsu9h3VlbJmrJDUBAu4BKuN3/pLUDfJRUlfIz3Sphma2WRsxYgT77LPPRqNCHnHEETzyyCONrlqpdegmpoi4OSKOyPOHRMTbImJkRHwqIlbm9IiIL0TE3jnf4wqYbcGam5u59957mTVrFq+88gqzZs3i3nvv9Q12PcyjQppZj/INdo3hgcPMzDZTbQ0c5iP3zVhnehv0hh9zM+t5Du6bsdYCtSQHcWuoznZz9H7bfRzczazb+cCj8Tzkr5lZCTm4m5mVkIO7mVkJObibmZWQg7uZWQk5uJuZlZCDu5lZCTm4m5mVkIO7mVkJObibmZWQg7uZWQk5uJuZlVDdwV1SH0l3S7ohv95L0h2SHpV0jaRtcnq//PrRnD+sZ6puZmat6ciR+z8DCwuvvwVcEBFvBpYB43P6eGBZTr8glzMzs02oruAuaSjwUeCS/FrAIcB1ucjlwFF5/sj8mpz/QXV2cGczM+uUeo/cvwv8K/Bqfr0zsDwi1uXXi4AheX4I8CRAzl+Ry29E0gRJ8yTNW7p0aSerb2ZmtbQb3CUdATwbEfO7c8MRcXFEjI6I0U1NTd25ajOzLV49T2J6L/AxSYcD/YEdgQuBgZK2zkfnQ4HFufxiYHdgkaStgTcAz3d7zc3MrFXtHrlHxFkRMTQihgHHA7Mj4kRgDnB0LnYy8PM8f31+Tc6fHX6ulpnZJtWVfu6TgNMlPUpqU5+e06cDO+f004Ezu1ZFMzPrqA49IDsibgZuzvN/Bg6sUWY1cEw31M3MzDrJd6iamZWQg7uZWQk5uJuZlZCDu5lZCTm4m5mVkIO7mVkJObibmZWQg7uZWQk5uJuZlZCDu5lZCTm4m5mVkIO7mVkJObibmZWQg7uZddqgQYOQVPcEdKi8JAYNGtTgd7l56tCQv2ZmRcuWLaOnn8VT+VGwjvGRu5lZCTm4m5mVULvBXVJ/SXdKulfSA5LOyemXSXpM0j15GpXTJel7kh6VdJ+kd/T0myizjrZpdqZd022aZuVTT5v7GuCQiFgpqS8wV9KsnPeViLiuqvw4YJ88vQv4Yf5rneA2TTPrjHaP3CNZmV/2zVNb0eZI4Iq83O3AQEm7db2qZmZWr7ra3CX1kXQP8Czwm4i4I2dNyU0vF0jql9OGAE8WFl+U06rXOUHSPEnzli5d2oW3YGZm1eoK7hGxPiJGAUOBAyWNBM4C9gUOAAYBkzqy4Yi4OCJGR8TopqamDlbbzMza0qHeMhGxHJgDfCQiluSmlzXAj4EDc7HFwO6FxYbmNDMz20Tq6S3TJGlgnt8W+DDwUKUdXelq3FHAgrzI9cCnc6+Zg4AVEbGkR2pvZmY11dNbZjfgckl9SD8G10bEDZJmS2oCBNwDnJrL/xI4HHgUeBn4TPdX28zM2tJucI+I+4D9a6Qf0kr5AL7Q9aqZmVln+Q5VM7MS8sBhZtZp8bUd4ew39Pw2rMMc3M2s03TOi5vkDuo4u0c3UUoO7r2cj4zMrDMc3Hs5HxmZWWf4gqqZWQk5uJuZlZCDu5lZCTm4m5mVkIO7mVkJObibmZWQg7uZWQk5uJuZlZBvYjKzLunpB6zvtNNOPbr+snJwN7NO6+jd05J6/I5rS9wsY2ZWQg7uZmYlVM8zVPtLulPSvZIekHROTt9L0h2SHpV0jaRtcnq//PrRnD+sZ9+CmZlVq+fIfQ1wSET8LTAK+Eh+8PW3gAsi4s3AMmB8Lj8eWJbTL8jlzMxsE2o3uEeyMr/sm6cADgGuy+mXA0fl+SPza3L+B9XTl9NLTlKPTu6NYFY+dfWWkdQHmA+8GfgB8CdgeUSsy0UWAUPy/BDgSYCIWCdpBbAz8FzVOicAEwD22GOPrr2LEutMzwL3SDCzui6oRsT6iBgFDAUOBPbt6oYj4uKIGB0Ro5uamrq6OjMzK+hQb5mIWA7MAd4NDJRUOfIfCizO84uB3QFy/huA57ultmZmVpd6ess0SRqY57cFPgwsJAX5o3Oxk4Gf5/nr82ty/uxwG4GZ2SZVT5v7bsDlud19K+DaiLhB0oPA1ZK+AdwNTM/lpwM/kfQo8AJwfA/U28zM2tBucI+I+4D9a6T/mdT+Xp2+GjimW2pnZmad4jtUzcxKyMHdzKyEHNzNzErIwd3MrIQc3M3MSsjB3cyshBzczcxKyMHdzKyEHNzNzErIwd3MrIQc3M3MSsjB3cyshBzczcxKyMHdzKyEHNzNzErIwd3MrIQc3M3MSqieZ6juLmmOpAclPSDpn3P62ZIWS7onT4cXljlL0qOSHpZ0WE++ATMze716nqG6DjgjIu6StAMwX9Jvct4FEXFesbCk/UjPTR0BDAZ+K+ktEbG+OytuZmata/fIPSKWRMRdef4lYCEwpI1FjgSujog1EfEY8Cg1nrVqZmY9p0Nt7pKGkR6WfUdO+qKk+yRdKmmnnDYEeLKw2CJq/BhImiBpnqR5S5cu7XDFzcysdXUHd0nbA/8NnBYRLwI/BPYGRgFLgP/syIYj4uKIGB0Ro5uamjqyqJmZtaOu4C6pLymwXxURPwWIiGciYn1EvApM47Wml8XA7oXFh+Y0MzPbROrpLSNgOrAwIs4vpO9WKPZxYEGevx44XlI/SXsB+wB3dl+Vzay3k1Rzaiuvkm/do57eMu8FTgLul3RPTpsMnCBpFBDA48DnASLiAUnXAg+Setp8wT1lzLYsEdHoKmzx2g3uETEXqPWT+ss2lpkCTOlCvczMrAt8h6qZWQk5uJtZj5s5cyYjR46kT58+jBw5kpkzZza6SqVXT5u7mVmnzZw5k+bmZqZPn86YMWOYO3cu48ePB+CEE05ocO3KS73hwsfo0aNj3rx5ja7GZqczvXL80gsAAAzxSURBVAt6w/dtW5aRI0fS0tLC2LFjN6TNmTOHiRMnsmDBgjaWtPZImh8Ro2vm9YZ/dgd3s/Lq06cPq1evpm/fvhvS1q5dS//+/Vm/3h3puqKt4O42dzPrUcOHD2fu3Lkbpc2dO5fhw4c3qEZbBgd3M+tRzc3NjB8/njlz5rB27VrmzJnD+PHjaW5ubnTVSs0XVM2sR1Uumk6cOJGFCxcyfPhwpkyZ4oupPcxt7mZmmym3uZtZQ7mf+6bnZhkz61Hu594YbpYxsx7lfu49x80yWwif+lpvtHDhQhYtWrTRvrlo0SIWLlzY6KqVmptlSsKnvtZbDR48mEmTJnHVVVdt2DdPPPFEBg8e3OiqlZqP3EtiypQpTJ8+nbFjx9K3b1/Gjh3L9OnTmTLFIy9b41U3//aG5uCyc5t7SfgWb+ut+vTpwyGHHMJNN91ERCCJD37wg8yePdv7Zhe5zX0L4Fu8rbcaOHAgs2fP5rzzzmPVqlWcd955zJ49m4EDBza6aqXWbpu7pN2BK4BdSY/UuzgiLpQ0CLgGGEZ6zN6xEbEsP3P1QuBw4GXglIi4q2eqbxXNzc0cd9xxDBgwgCeeeII999yTVatWceGFFza6araFe/HFFxk4cCD7778/ffv2Zf/992fgwIG8+OKLja5aqdVz5L4OOCMi9gMOAr4gaT/gTOCmiNgHuCm/BhhHeij2PsAE4IfdXmtrkx80bL3JunXrOProoxk3bhzbbLMN48aN4+ijj2bdunWNrlqptRvcI2JJ5cg7Il4CFgJDgCOBy3Oxy4Gj8vyRwBWR3A4MlLRbt9fcNjJlyhSuueYaHnvsMdavX89jjz3GNddc4wuq1nBbb7011113HbNmzeKVV15h1qxZXHfddWy9tTvr9aQOtblLGgbsD9wB7BoRS3LW06RmG0iB/8nCYotymvWghQsXMmbMmI3SxowZ477E1nA77rgjK1as4O6772bt2rXcfffdrFixgh133LHRVSu1uoO7pO2B/wZOi4iNGssidbnpULcbSRMkzZM0b+nSpR1Z1GrwBVXrrZYvX86ECROYPHkyAwYMYPLkyUyYMIHly5c3umqlVldwl9SXFNivioif5uRnKs0t+e+zOX0xsHth8aE5bSMRcXFEjI6I0U1NTZ2tv2UeM9t6q+HDh3PMMcewevVqIoLVq1dzzDHH+MCjp0VEmxMgUm+Z71alfwc4M8+fCXw7z38UmJWXOwi4s71tvPOd7wzruhkzZsSIESNiq622ihEjRsSMGTMaXSWzmDFjRuy1114xe/bseOWVV2L27Nmx1157ef/sBsC8aCWutnsTk6QxwK3A/cCrOXkyqd39WmAP4AlSV8gXclfI7wMfIXWF/ExEtHmHkm9iMiu3mTNnMmXKlA0P62hubvawGN3AD8g2Mysh36FqZraFcXA3MyshB3czsxJycDczKyEHdzOzEnJwNzMrIQd3M7MScnA3sx7nh7dveh5z08x6lB/e3hi+Q9XMetTIkSNpaWlh7NixG9LmzJnDxIkTWbBgQQNrtvnz8ANm1jB+eHvP8fADWwi3a1pv5GcNNIaDe0lU2jVbWlpYvXo1LS0tNDc3O8Bbw/lZAw3S2ljAm3LyeO5dN2LEiJg9e/ZGabNnz44RI0Y0qEZmr/GzBnoGXRnPfVNwm3vXuV3TbMvjNvctgNs1zazIwb0k3K5pZkW+iakkKjeDTJw4ccOjzKZMmeKbRMy2UPU8Q/VS4Ajg2YgYmdPOBj4HLM3FJkfEL3PeWcB4YD3wpYj4dXuVcJu7mVnHdbXN/TLSw66rXRARo/JUCez7AccDI/IyF0nq07lqm5lZZ7Ub3CPiFuCFOtd3JHB1RKyJiMeAR4EDu1A/MzPrhK5cUP2ipPskXSppp5w2BHiyUGZRTnsdSRMkzZM0b+nSpbWKmJlZJ3U2uP8Q2BsYBSwB/rOjK4iIiyNidESMbmpq6mQ1zMyslk4F94h4JiLWR8SrwDRea3pZDOxeKDo0p5mZ2SbUqa6QknaLiCX55ceByrid1wMzJJ0PDAb2Ae5sb33z589/TtITnamL1bQL8FyjK2FWg/fN7rVnaxntBndJM4GDgV0kLQK+BhwsaRQQwOPA5wEi4gFJ1wIPAuuAL0REu/e+R4TbZbqRpHmtdY8yayTvm5tOrxhbxrqX/4Gst/K+uel4+AEzsxJycC+nixtdAbNWeN/cRNwsY2ZWQj5yNzMrIQd3M7MScnBvIEmPS9ql0fXYlCSdJmm7RtdjcyBpmKQFreT9u6QPtbP82ZK+3DO1A0mDJV1XR7mVraQflQcb3KxIGijpnxpdj/Y4uHeCEn92nXMa4ODeRRHx1Yj4bYPr8FREHN2FVRwFbHbBHRgIOLiXRT6KeljSFaQ7cneX9MM8+NkDks4plH1c0jmS7pJ0v6R9c/rOkm7M5S8BVFjmdEkL8nRaYZsPSbpM0h8lXSXpQ5Juk/SIpNeNuClpO0nXSnpQ0s8k3SFpdM47VNLvc73+S9L27dR3QB4Y7k5Jd0s6ssb2dpN0i6R7ct3f19q2JH2JdOfyHElzuu3LKbc+kqblfeZGSdsC5H3i6Dx/eN5P5kv6nqQbCsvvJ+lmSX/On//r5O98YD5oeV7Sp3P6FZI+LKmPpO9I+oPSYIGfz/kbziza2u9y/hRJ90q6XdKukt4DfAz4Tt539q6q09657P2SvlE8+pf0lUJdzinUZWErn9Xekn6VP59bK/t31fY+kOtxT97Xd2htW8A3gb1z2e904LvctFp7cranjSdgGPAqcFAhbVD+2we4GXh7fv04MDHP/xNwSZ7/HvDVPP9R0h2+uwDvBO4HBgDbAw8A++dtrgPeRvohng9cSvpROBL4nxr1/DLwozw/Mi8/Om/nFmBAzptUqEtr9T0X+FSeHwj8sbJ8YXtnAM2Fz2GHOra1S6O/z81hKnz/o/Lrawvfx2XA0UB/0kise+X0mcANef5s4P+Afvk7eR7oW2M7U/P+OBL4AzAtpz+S98kJwL/ltH7APGCvXL8Fbe13+XUAf5fnv11Y12XA0a289xuAE/L8qcDKPH8oqTul8v/EDcD72/msbgL2yfPvAmbX2N7/Au/N89uT7t5va1sLGr1/tDf5yL1jnoiI2wuvj5V0F3A36QElxVPMn+a/80k7A6Qd40qAiPgFsCynjwF+FhGrImJlXvZ9Oe+xiLg/0iBtDwA3RdoD7y+st2gMcHXexgLgvpx+UK7fbZLuAU5m43EpatX3UODMXP5mUiDZo2p7fwA+o/R0rrdFxEt1bMvq91hE3JPni99Nxb7AnyM9PwFScC/6RaTnKzwHPAvsWmMbt5L2zfeTRnx9m6QhwLKIWEXaDz6dv8s7gJ1J40YVtbbfAbxCCoytvYda3g38V56fUUg/NE93A3eR3n+lLq/7rPLZ6XuA/8r1/xGwW43t3Qacn89uBkbEuna21ev5Gaods6oyI2kv0tHKARGxTNJlpOBXsSb/XU/XPuc1hflXC69f7eB6BfwmIlp7qGqt+gr4REQ83NpKI+IWSe8nHfldpjRo3LJ2tmX1K37/64Ftu7j81pK+QHpMJsDhpLOsL5B+uJtJgwEeTQr6kPaDiVH1yExJw+qsw9p8QLKhDh17CxsR8B8R8aMadan1WW0FLI+IUW2tNCK+KekXpM/jNkmHtbOtXs9H7p23IynYr5C0KzCujmVuAT4JIGkcUHnIya3AUbndcgDpn+vW2qto123AsXkb+5GadABuB94r6c05b4Ckt7Szrl8DEyUpL7N/dQFJewLPRMQ04BLgHe1s6yVS0411j4eBNxUCznHtLRARP4jXHpH5VEQ8SWq22Sci/gzMJR243JIX+TXwj5L6Akh6S95Pi1rb79rS1r5wO/CJPH98If3XwGf12vWiIZLe2MZ7fRF4TNIxubwk/W11OUl75zPkb5HORvdtY1ubxT7s4N5JEXEv6XTtIdJp4211LHYO8H5JDwB/D/wlr+suUvvjnaTT3ksi4u5OVu0ioEnSg8A3SE05KyJiKXAKMFPSfcDvSTtwW74O9AXuy3X+eo0yBwP3SrqbFFgubGdbFwO/ki+odouI+CvpOsmvJM0nBZ4VnVjVHaRrKpAOLIaQgjykH+0HgbvyBdQf8fqj75r7XTvbvBr4Sr6AuXdV3mnA6Xn/eXNlXRFxI+n/7feS7geuo/1AeyIwXtK9uV6v6xgAnKbUIeA+YC0wq7VtRcTzpKP7Bb35gqqHHygZpQeS942I1fkf5rfAWyPilQZXzXqIpO0jYmU+w/oB8EhEXLCJ69Ct+53SvRB/jYiQdDzp4mqtoGytcJt7+WxH6mrYl9Rm+E8O7KX3OUknA9uQziZ/1E75ntDd+907ge/nH6zlwGe7oY5bFB+5m5mVkNvczcxKyMHdzKyEHNzNzErIwd3MrIQc3M3MSuj/AzZJbkmBJHx6AAAAAElFTkSuQmCC\n",
            "text/plain": [
              "<Figure size 432x288 with 1 Axes>"
            ]
          },
          "metadata": {
            "tags": [],
            "needs_background": "light"
          }
        }
      ]
    }
  ]
}