R/Utility.R

In JBrownBiostat/Dino: Normalization of Single-Cell mRNA Sequencing Data

# calcDepthRep is a helper function to calculate the replications in the depth
# vector. Procedure is close to that of table, but maintains the same rounding
# scheme as other functions
calcDepthRep <- function(depth) {
    depth <- sort(depth)
    dupDepth <- duplicated(depth)
    freq <- c(which(!dupDepth), length(depth) + 1)
    return(data.frame(
        depth = depth[!dupDepth],
        rep = freq[-1] - freq[-length(freq)]
    ))
}


# setPar is a helper function to initiate the parallel environment
#
#' @importFrom BiocParallel SnowParam
#' @importFrom BiocParallel MulticoreParam
#' @importFrom BiocParallel register
setPar <- function(nCores, taskList = NULL) {
    if(is.null(taskList)) {
        nTask = 0L
    } else {
        nTask <- ceiling(sqrt(length(taskList) / nCores)) * nCores
    }
    if (.Platform$OS.type == "windows") {
        prll <- SnowParam(workers = nCores, tasks = nTask, progressbar = TRUE)
        register(BPPARAM = prll, default = TRUE)
    } else {
        prll <- MulticoreParam(workers = nCores, tasks = nTask,
                    progressbar = TRUE)
        register(BPPARAM = prll, default = TRUE)
    }
    return(prll)
}


# splitGenes is a helper function to split a sparse count matrix into a list of
# gene sets for parallel computation
#
#' @importFrom Matrix Matrix
splitGenes <- function(counts, nCol = 40) {
    if(nCol >= ncol(counts)) {
        return(list(counts))
    }
    m <- floor(ncol(counts) / nCol)
    ret <- lapply(seq_len(m), function(i) {
        counts[, (nCol * i - (nCol - 1)):(nCol * i), drop = FALSE]
    })
    if(m * nCol < ncol(counts)) {
        ret[[m + 1]] <- counts[, (nCol * m + 1):ncol(counts), drop = FALSE]
    }
    return(ret)
}