Task21_Collection_Key_Results.Rmd

---
title: "Task21 - Key Examples, perhaps for discussion"
author: Andrew Stein
output:
  html_document:
    toc: true
    toc_float: true
    code_folding: hide
---
# Introduction
The idea is to identify cases where AFIR theory doesn't match SCIM simulation and understand why and build in tests for this.

# Setup and Read Data
```{r, warning=FALSE, message=FALSE}
source("ams_initialize_script.R")
source("SCIM_calculation.R")  
source("ivsc_2cmt_RR_V1.R")
library(RxODE)
model = ivsc_2cmt_RR_KdT0L0()
dirs$rscript_name = "Task21_Collection_Key_Results.Rmd"
dirs$filename_prefix= str_extract(dirs$rscript_name,"^Task\\d\\d\\w?_")

data_in = read.csv("results/Task20_data.csv",stringsAsFactors = FALSE)

data = data_in %>%
  mutate(target = ifelse(is_soluble==1,"soluble","membrane"))

assumptions = data %>%
  select(id,AFIR_thy,AFIR_sim,SCIM_simplest_thy,SCIM_adhoc_thy,SCIM_sim,starts_with("assumption")) %>%
  arrange(SCIM_sim)
nam = names(assumptions) %>%
  str_replace("^assumption_","")
names(assumptions) = nam
```
# Explore agreement between AFIR theory and SCIM simulation

List of assumptions

 * AFIR < 0.30,
 * Cavgss >> Ttotss (5x larger)
 * Cavgss >> Kss_DT (5x)
 * koff_DT > keT (binding faster than elimination)
 * koff_TL > 1/30 (not too slow)
 * Cavgss >> Kss_DT*Lss/Kss_TL (enough binding)
 * Lss/L0 ~ 1 (no accumulation of ligand)

```{r, warning = FALSE, message = FALSE}
data_plot = data

g = ggplot(data_plot, aes(AFIR_SCIM_pcterr*100, fill = assumption_all_AFIR))
g = g + geom_histogram()
g = g + xgx_scale_x_log10()
g = g + facet_wrap(~target)
g = g + scale_fill_manual(values = c(`TRUE`="grey50",`FALSE`="pink"))
g = g + labs(x = "Percent Error",
             y = "Number of Simulations")
g = g + ggtitle("AFIR theory vs\nSCIM simulation")
print(g)
```


# Explore agreemnet between SCIM theory and SCIM simulation

List of assumptions.  They are similar to the AFIR assumptions, but now we allow for accumulation of Ligand

 * SCIM < 30%
 * Cavgss >> Ttotss (5x larger)
 * Cavgss >> Kss_DT (5x)
 * koff_DT > keT (binding faster than elimination)
 * koff_TL > 1/30 (not too slow)
 * Cavgss >> Kss_DT*Lss/Kss_TL (enough binding)

```{r, warning = FALSE, message = FALSE}
g = ggplot(data_plot, aes(SCIM_SCIM_pcterr*100, fill = assumption_all_SCIM))
g = g + geom_histogram()
g = g + xgx_scale_x_log10()
g = g + facet_wrap(~target)
g = g + scale_fill_manual(values = c(`TRUE`="grey50",`FALSE`="pink"))
g = g + labs(x = "Percent Error",
             y = "Number of Simulations")
g = g + ggtitle("SCIM adhoc Lfold theory vs\nSCIM simulation")
print(g)
```

# Look into large error with assumptions - check tol for ODE

```{r, warning = FALSE, message = FALSE}
data_focus = data %>%
  filter(assumption_all_SCIM == TRUE) %>%
  arrange(desc(SCIM_SCIM_pcterr)) %>%
  mutate(Dss_TLss_thy_ratio = Dss_thy/TLss_thy,
         Dss_TLss_sim_ratio = Dss_sim/TLss_sim)

data_focus %>%
  select(id,SCIM_SCIM_pcterr,contains("Dss_TLss")) %>%
  slice(1:10) %>%
  kable()
```

# D is way lower than predicted
### Because Dss < Ccrit, so all the Drug is getting internalized
From the theory:

 * Ttotss = 100
 * Dss = 600
 * keDT = 13
 * CL = .27
 * V  = 3
 * ...
 * linear elimination    = Dss*CL = 600*.27 ~ 150
 * nonlinear elimination = 
 * Ccrit = Vm/CL = ksynT*V1/CL = 1375*3/.27 = 15,000
 
 Ok, so Dss is not large enough.  We need a check that it's above Ccrit.
 
 Shoot, I forgot what run illustrated this issue...  It's not 9129.  I think it's a mistake that I plot it below.
 
```{r, warning = FALSE, message = FALSE}
id_plot = 9129
d = data %>% filter(id==id_plot)

out = plot_param(d,model)
kable(out$param)
kable(out$compare)
```

# This simulation seems to be giving a negative TLss
### This is due to the large difference between Dss and TLss
ODE can't solve it accurately.  But only one patient like this (see above)

```{r, warning = FALSE, message = FALSE}
id_plot = 7651
d = data %>% filter(id==id_plot)

print(paste("Cavgss: ", d$Cavgss))
print(paste("TLss: ", d$TLss_thy))

out = plot_param(d,model, plot_flag = FALSE)
print(out$plot)
kable(out$param)
kable(out$compare)
print(d$SCIM_sim)
```


# Look at a patient with all assumptions true and large error

### Something strange is going on, with the Dss_sim and Dss_thy
```{r, warning = FALSE, message = FALSE}
x = data_focus
d = x[2,]

out = plot_param(d,model)
kable(out$param)
kable(out$compare)
print("original simulation result")

Dss_calc = with(d,dose_nmol/CL/tau)
d %>%
  select(id,Dss_sim, Dss_thy, T0_sim, L0_sim, TL0_sim, Ttotss_sim, Lss_sim, TLss_sim, AFIR_sim, SCIM_sim, SCIM_Lfold_adhoc_thy) %>%
  kable()
```

# Continue looking at patient with all assumptions true and large error
```{r, warning = FALSE, message = FALSE}
out = plot_param(x[3,],model)
kable(out$param)
kable(out$compare)

out = plot_param(x[4,],model)
kable(out$param)
kable(out$compare)

out = plot_param(x[5,],model)
kable(out$param)
kable(out$compare)
```