Do we really need BiocParallel for generateNull()?

[MomenehForoutan]

I have been getting some errors using bplapplyfrom this package; something similar to what is raised up here "https://support.bioconductor.org/p/92587/". Would it be a problem if we use ldply() instead of bplapply(); here is what I would suggest: it basically gave the same results and it was even faster! I used that within a foreachloop though... Can you please check this on the example data?

generateNull <- function(n_up, n_down, rankData, B = 1000, seed = 1){
  all_genes <- rownames(rankData)
  totalNo <- n_up + n_down

  set.seed(seed)

  temSets <- lapply(1:B, function(x) {
    sample(all_genes, size = totalNo, replace = FALSE)
  })
  r <- ldply(1:B, function(x) {
    tms <- temSets[[x]]

    if (n_down > 0) {
      upSet <-  GeneSet(as.character(tms[1:n_up]))
      downSet <-  GeneSet(as.character(tms[-(1:n_up)]))
      ss = simpleScore(rankData, upSet = upSet, downSet = downSet)
    } else {
      #else all the random generated genes are in upSet
      ss = simpleScore(rankData, upSet = GeneSet(as.character(tms)))
    }
    ss[, 1]
  })
  colnames(r) <- colnames(rankData)
  return(r)
}

[ruqianl]

Hi Sepideh,

I just had a test run of the two versions (with and without BiocParallel), it was actually faster when using BiocParallel on my Mac Book.

Time difference of 36.55549 secs vs. Time difference of 1.126696 mins

I haven't tested on Windows, but I will look into that. Thanks!

[ruqianl]

I figured out what's causing that and I found a solution for that