memento

.pre-commit-config.yaml

@@ -82,10 +82,14 @@ repos:
         exclude: ^tools/(cellxgene_census_builder|census_contrib)
         args: ["--config", "./tools/pyproject.toml"]
         additional_dependencies:
+          - attrs
           - numpy
           - pandas-stubs
           - typing_extensions
           - types-PyYAML
+          - pytest
+          - types-click
+
 
   - repo: https://github.com/igorshubovych/markdownlint-cli
     rev: v0.39.0

.vscode/launch.json

@@ -0,0 +1,72 @@
+{
+    "version": "0.2.0",
+    "configurations": [
+        {
+            "name": "Memento (tiny)",
+            "type": "python",
+            "request": "launch",
+            "cwd": "${workspaceFolder}/api/python/cellxgene_census/src/",
+            "module": "cellxgene_census.experimental.diffexp.memento.diff_expr",
+            "justMyCode": true,
+            "args": [
+                "tissue_general_ontology_term_id in ['UBERON:0002405']", 
+                "sex_ontology_term_id",
+                "/mnt/census/estimators-cube-70a7705/",
+                "1",
+                "1000"],
+            "subProcess": true
+        },
+        {
+            "name": "Memento (small)",
+            "type": "python",
+            "request": "launch",
+            "cwd": "${workspaceFolder}/api/python/cellxgene_census/src/",
+            "module": "cellxgene_census.experimental.diffexp.memento.diff_expr",
+            "justMyCode": true,
+            "args": [
+                "tissue_general_ontology_term_id in ['UBERON:0000030', 'UBERON:0000992']", 
+                "tissue_general_ontology_term_id",
+                "/mnt/census/estimators-cube-70a7705/",
+                "1",
+                "1000"],
+            "subProcess": true
+        },
+        {
+            "name": "Memento (medium)",
+            "type": "python",
+            "request": "launch",
+            "cwd": "${workspaceFolder}/api/python/cellxgene_census/src/",
+            "module": "cellxgene_census.experimental.diffexp.memento.diff_expr",
+            "justMyCode": true,
+            "args": [
+                "tissue_general_ontology_term_id in ['UBERON:0000948', 'UBERON:0001004']", 
+                "tissue_general_ontology_term_id",
+                "/mnt/census/estimators-cube-70a7705/",
+                "1",
+                "5000"],
+            "subProcess": true
+        },        
+        {
+            "name": "Memento (large)",
+            "type": "python",
+            "request": "launch",
+            "cwd": "${workspaceFolder}/api/python/cellxgene_census/src/",
+            "module": "cellxgene_census.experimental.diffexp.memento.diff_expr",
+            "justMyCode": true,
+            "args": [
+                "tissue_general_ontology_term_id in ['UBERON:0000948', 'UBERON:0001004']", 
+                "tissue_general_ontology_term_id",
+                "/mnt/census/estimators-cube-70a7705/",
+                "1",
+                ""],
+            "subProcess": true
+        },        
+        {
+            "name": "Python: File",
+            "type": "python",
+            "request": "launch",
+            "program": "${file}",
+            "justMyCode": true
+        }
+    ]
+}

api/python/cellxgene_census/pyproject.toml

@@ -45,6 +45,7 @@ experimental = [
     "torchdata~=0.7",
     "scikit-learn~=1.0",
     "scikit-misc>=0.2",  # scikit-misc 0.3 dropped Python 3.8 support
+    "polars==0.20.4",
     "psutil~=5.0",
     "datasets~=2.0",
     "tdigest~=0.5",

api/python/cellxgene_census/src/cellxgene_census/experimental/diffexp/memento/README.md

@@ -0,0 +1,9 @@
+# Differential Expression using memento
+
+This directory contains code for a Census-integrated version of the `memento` method for differential expression
+analysis, including differential variability and co-expression. The underlying method is described in
+the [memento pre-print](https://www.biorxiv.org/content/10.1101/2022.11.09.515836v1).
+
+This implementation relies upon a database of pre-computed estimators that are derived from a given Census data release.
+The database is a TileDB array, structured as a multi-dimensional cube. It is built by
+the `tools/models/memento/src/estimators_cube_builder/cube_builder.py` script.