R/read_ipa.R

In HuntsmanCancerInstitute/hciR: RNA-seq workflows at HCI

#' Read IPA Expression Analysis
#'
#' Read an IPA Expression Analysis export all and optionally
#' write to an  Excel file with each table in a separate tab.
#'
#' @param file IPA output file (created using Export All on the main Summary tab)
#' @param excel write to Excel file
#' @param mylists include optional My Pathways and My Lists
#'
#' @return a list of tibbles or Excel file
#'
#' @author Chris Stubben
#'
#' @examples
#'  \dontrun{
#'  x <- read_ipa("RPR2_vs_RPM.txt")
#'  names(x)
#'  read_ipa("RPR2_vs_RPM.txt",  excel=TRUE)
#'  }
#' @export

read_ipa <- function(file,  excel=FALSE, mylists=FALSE){
   x <- readr::read_lines(file)
   ## Find start of tables
   n <- grep("for My Projects", x)
   # save results
   n1 <- length(n)
   ## check...
   if(n1 ==0) stop("No My Projects headers. File should be an IPA output file from Export All")
   z <- vector("list", n1)
   ## get names for Excel tabs
   TABS <-  gsub( " for My .*", "", x[n])
   names(z) <- TABS
   ## add length of x
   n <- c(n, length(x)+1)
   # Loop through tables...
   for(i in 1:n1){
      start <- n[i] + 1
      end <- n[i + 1] - 2
      if(!mylists) if( TABS[i]  %in% c("My Lists", "My Pathways") ) next
      if(end - start == 0) next
      message("Loading ", TABS[i])
      y <- x[  start:end  ]
      ## fix 1 row with  3 columns
      if(i == 1)  y <- gsub("\tand consider Both up", " and consider Both up", y)
	  ## first table with Analysis Details returns Warning: 5 parsing failures.
      z[[i]] <- suppressWarnings( readr::read_tsv( paste( gsub("\t$", "", y), collapse ="\n"), show_col_types = FALSE) )
    }
    ## fix Pathways, rename columns, add total matches, set size

    z[["Canonical Pathways"]] <- dplyr::rename(z[["Canonical Pathways"]],
          Pathway =`Ingenuity Canonical Pathways`, pValue = `-log(p-value)`) %>%
       dplyr::mutate(pValue = 10^(-pValue),
         #  zScore = as.numeric(sprintf("%.3f", zScore)),
		  zScore = round(zScore, 3),
               N = nchar( gsub("[^,]" , "", Molecules))+1 ,
         setSize = round( N / Ratio, 0)) %>%
       dplyr::arrange(pValue)
    ## NaN read as #NUM! in Excel
    z[["Canonical Pathways"]]$zScore[ is.nan( z[["Canonical Pathways"]]$zScore)] <- NA
    ## add adjusted p-value
    z[["Canonical Pathways"]]$padj <- p.adjust(z[["Canonical Pathways"]]$pValue, method="BH")

    # drop Analysis name and empty ID
    z[["Upstream Regulators"]] <- z[["Upstream Regulators"]][,-(1:2)]
    z[["Networks"]] <-  z[["Networks"]][,-2]
    # drop missing tables , usually My lists
    z <- z[!sapply(z, is.null)]
    if(excel){
      z <- lapply(z, data.frame, check.names=FALSE)
      outfile <- gsub(".txt", ".xlsx", file)
      message("Saved to ", outfile)
      openxlsx::write.xlsx(z, file = outfile )
    }else{
      z
    }
}