R/GetBarcodes.R

In noush-joglekar/scisorseqr: Processes long reads for differential isoform analysis

#' Get single-cell barcodes from long read files
#'
#' @description This is an R-wrapper for a python function. It parallelizes the reading
#' and processing of fastq.gz files, and uses a sliding window approach to
#' identify cell barcodes, assign cluster labels and output useful statistics
#' in one .csv file and one _summary file per fastq.gz input file
#' @param fqFolder Path to a folder containing fastq.gz files
#' @param BCClustAssignFile tab separated file containing barcode in column 1
#' and cluster in column 2
#' @param outputFolder OPTIONAL Path to working directory. Defaults to current
#' @param numProcesses Number of chunks to split processing into. Defaults to 10
#' @param chemistry UMI length differs with 10x chemistry. Defaults to "v2"
#' @param concatenate OPTIONAL Concatenate output from all files in the folder
#' @param filterReads OPTIONAL logical indicating whether the barcoded reads should be
#' filtered into a separate file. Will make the downstream analysis faster if you expect
#' few barcoded reads (~40 of total) at the cost of a slow (single-threaded) filtering process
#' @return OutputRaw folder containing a .csv file and summary stats for each
#' input fastq.gz file
#' @return OutputFiltered Filtered file with one line for each read containing
#' a barcode. Option to concatenate into one file per input folder
#' @usage GetBarcodes("FastqFolder","Barcode-Clust","~/MyDir/",6,
#' concatenate = FALSE,filterReads = FALSE)
#' @export
GetBarcodes <- function(fqFolder, BCClustAssignFile, outputFolder = ".", numProcesses = 10,
                        chemistry = "v2", concatenate = TRUE, filterReads = FALSE) {

  if(!dir.exists(outputFolder)){dir.create(outputFolder)}

  raw_Output <- paste0(outputFolder, "/OutputRaw/")

  py_file <- system.file("python", "BarcodeDeconvolution.py", package = "scisorseqr")
  simConcat <- system.file("bash", "concat_ParallelizedBC.sh", package = "scisorseqr")
  filterOut <- system.file("bash", "FilterBCReads.sh", package = "scisorseqr")

  # Creates subdirectory for unfiltered output and summary files
  if(!dir.exists(raw_Output)){dir.create(raw_Output)}
  files <- list.files(fqFolder)[grep(pattern = "q.gz",list.files(fqFolder))]

  # Iterates through each file, activating necessary scripts for
  # analysis and concatenation of parallel processes
  for (file in files) {
    input_file <- paste0(fqFolder, file)
    f_name <- unlist(strsplit(file,".fastq|.fq"))[1]
    run_bc_deConv <- paste("python3", py_file, input_file, BCClustAssignFile,
                           "--outDir", raw_Output, "--numProc", numProcesses,
                           "--chemistry",chemistry)
    run_concatSimProc <- paste("bash", simConcat, f_name, raw_Output)
    system(run_bc_deConv)
    system(run_concatSimProc)
  }

  # Calls next function automatically once all files have been analyzed and
  # .csv files created.
  scisorseqr::FilterBCOutput(outputFolder, raw_Output, concatenate, filterReads, fqFolder)
}