SCOUT scripts can only be executed on Orfeo

[albertocasagrande]

All SCOUT scripts adopt hardcoded absolute paths that depend on the machine used to produce the data, i.e., Orfeo.

For instance, SPN01 simulate_mutations.R script contains the following lines.

outdir <- "/orfeo/scratch/cdslab/shared/SCOUT/SPN01/races/"
forest <- load_samples_forest(paste0(outdir,"sample_forest.sff"))


setwd("/orfeo/cephfs/scratch/cdslab/shared/ProCESS/GRCh38")
m_engine <- MutationEngine(setup_code = "GRCh38",tumour_type = "COAD",
                           tumour_study = "US")

This approach works on Orfeo, but it is not general at all and prevents using SCOUT scripts on different machines or directory configurations. This is a major issue because data reproducibility is one of the main goal of SCOUT.

I suggest removing the absolute paths and exclusively accessing the subdirectories of the working directory. For instance, the above line would become

outdir <- "output"
forest <- load_samples_forest(file.path(outdir, "sample_forest.sff"))

m_engine <- MutationEngine(setup_code = "GRCh38",tumour_type = "COAD",
                           tumour_study = "US")

If using subdirectories is not optimal, for instance because you want to share the same mutation engine directory, you can either

• define the subdirectories "output" and "GRCh38" as symbolic links by executing the following command-line lines

   ln -s  /orfeo/scratch/cdslab/shared/SCOUT/SPN01/races/ output
   ln -s /orfeo/cephfs/scratch/cdslab/shared/ProCESS/GRCh38/GRCh38 GRCh38
  

• add two parameters to the SCOUT scripts to explicity get output and mutation engine directories. In this case, the original code snippet could become:

   args <- commandArgs(trailingOnly = TRUE)
   if (length(args) != 2) {
       args <- commandArgs(trailingOnly = FALSE)
       script_path <- sub("^--file=", "", args[grep("^--file=", args)])
   
       stop(paste("Syntax error: Rscript", basename(script_path), "<output_dir> <mutation_engine_directory>"))
   }
   
   output <- args[1]
   mu_dir <- args[1]
   forest <- load_samples_forest(paste0(outdir,"sample_forest.sff"))
  
   setwd(mu_dir)
   m_engine <- MutationEngine(setup_code = "GRCh38",tumour_type = "COAD",
                          tumour_study = "US")
  

Using paste0() to join paths should also be deprecated in favor of file.path().

[giorgiagandolfi]

Yes we know, at the moment we prefer having Orfeo specific configuration in order to standardise all analyses (and be sure that everyone run everything in the same way.) As soon as everyone finish to run the sequencing we will modify the repository in order to be independent from Orfeo.