data_preprocessing.Rmd

---
title: "Ensemble Boolean Model Data Preprocessing"
author: "[John Zobolas](https://github.com/bblodfon)"
output: 
  html_document:
    theme: united
---

## Intro

The purpose of this R markdown document is to extract from one or multiple 
**models** directories the models' **stable states** and **boolean link 
operators** to proper R data structures and save them for later use. This data 
extraction can take some time and it is better to have the data in tab-delimited
files, ready to be loaded as matrices before the analysis is done for each cell 
line.

## Data Structure and Output

The structure of the data and input to this document should be a 
single directory (`input.dir`). This directory must have
one or many `cell line` directories, each one having:

- A **models** directory (the models produced by the `Gitsbe` module in 
`.gitsbe` format).
- A **model_predictions** file
- An **observed_synergies** file 
- Maybe a **training_data** or **steady_state** file!

The output will be two more files in each cell line directory: **models_stable_state** 
and **models_equations** or **models_link_operator**.

**Note** that if instead of a `models` directory a compressed file `models.tar.gz` is there, you will have to run the command `tar xzvf models.tar.gz`.

## Data extraction

Necessary libraries:
```{r Load libraries, message = FALSE}
library(usefun)
library(emba)
```

For input, we provide the relative path to the data dir:
```{r Input}
#input.dir = "/atopo/cell-lines-2500"
input.dir = "/cascade/cell-lines-2500"
data.dir = paste0(getwd(), input.dir)

cell.line.dirs = list.dirs(path = data.dir, recursive = FALSE)
```

Note that below R code does not subset the models to the **unique** ones 
(those that have the exactly the same boolean equations, i.e. same link 
operators in our models). Such a processing should be done (if it is desirable) 
when the models' data is loaded in a seperate analysis document.

```{r Data extraction and saving, eval=FALSE}
for (cell.line.dir in cell.line.dirs) {
  # cell line name same as the directory name
  cell.line = basename(cell.line.dir)
  print(paste0("Saving files for ", cell.line, " cell line..."))
  models.dir = paste0(cell.line.dir, "/models")
  
  model.predictions.file = paste0(cell.line.dir, "/model_predictions")
  models.stable.state.file = paste0(cell.line.dir, "/models_stable_state")
  #models.equations.file = paste0(cell.line.dir, "/models_equations")
  models.equations.file = paste0(cell.line.dir, "/models_link_operator")
  
  if (file.exists(model.predictions.file)) {
    # for having the row names in the same order as in the 'model_predictions' file
    model.predictions = get_model_predictions(model.predictions.file)
    models = rownames(model.predictions)
  
    models.stable.state = get_stable_state_from_models_dir(models.dir)
    models.stable.state = models.stable.state[models,]
    save_mat_to_file(mat = models.stable.state, file = models.stable.state.file)
  
    models.equations = get_link_operators_from_models_dir(models.dir)
    models.equations = models.equations[models,]
    save_mat_to_file(mat = models.equations, file = models.equations.file) 
  }
}
```