biocParallel学习

我好像做了一个愚蠢的测试

rm(list=ls())
suppressPackageStartupMessages({library(SingleCellExperiment)library(scMerge)library(scater)library(Matrix)
})setwd("/Users/yxk/Desktop/test/R_parallel/")
load("./data/exprsMat.RData")
load("./data/clust.RData")
load("./data/pseudobulk_sample_list.RData")
load("./data/pseudobulk_sample.RData")
load("./data/use_bpparm.RData")
load("./data/res.RData")#' @importFrom ruv replicate.matrix
#' @importFrom methods as isaggregate.Matrix <- function(x, groupings=NULL) {if (!methods::is(x,'Matrix')) {x <- methods::as(as.matrix(x), "CsparseMatrix")}groupings2 <- paste("A", groupings, sep = "")if (length(unique(groupings2)) > 1) {mapping <- methods::as(ruv::replicate.matrix(groupings2), "CsparseMatrix")colnames(mapping) <- substring(colnames(mapping), 2)mapping <- mapping[, levels(factor(groupings))]} else {mapping <- methods::as(matrix(rep(1, length(groupings2)), ncol = 1), "CsparseMatrix")colnames(mapping) <- unique(groupings)}result <- t(mapping) %*% xreturn(result)
}create_pseudoBulk_parallel = function (exprsMat, cell_info, k_fold = 30, use_bpparam = BiocParallel::SerialParam()) 
{#browser()k_fold <- min(ncol(exprsMat), k_fold)cv <- cvTools::cvFolds(ncol(exprsMat), K = k_fold)exprsMat_pseudo <- BiocParallel::bplapply(seq_len(k_fold), function(i) {subset_idx <- cv$subsets[cv$which == i]cellType_tab <- table(droplevels(factor(cell_info[subset_idx])))cellTypes_n_mat <- matrix(rep(cellType_tab, nrow(exprsMat)), nrow = length(cellType_tab), byrow = FALSE)rownames(cellTypes_n_mat) <- names(cellType_tab)res <- aggregate.Matrix(t(exprsMat[, subset_idx]), cell_info[subset_idx])cellTypes_n_mat <- cellTypes_n_mat[rownames(res), ]res <- res/cellTypes_n_matrownames(res) <- paste(rownames(res), i, sep = "_")res}, BPPARAM = use_bpparam)exprsMat_pseudo <- do.call(rbind, exprsMat_pseudo)return(exprsMat_pseudo)
}create_pseudoBulk_no = function (exprsMat, cell_info, k_fold = 30) 
{#browser()k_fold <- min(ncol(exprsMat), k_fold)cv <- cvTools::cvFolds(ncol(exprsMat), K = k_fold)exprsMat_pseudo =list()for (i in seq_len(k_fold)){subset_idx <- cv$subsets[cv$which == i]cellType_tab <- table(droplevels(factor(cell_info[subset_idx])))cellTypes_n_mat <- matrix(rep(cellType_tab, nrow(exprsMat)), nrow = length(cellType_tab), byrow = FALSE)rownames(cellTypes_n_mat) <- names(cellType_tab)res <- aggregate.Matrix(t(exprsMat[, subset_idx]), cell_info[subset_idx])cellTypes_n_mat <- cellTypes_n_mat[rownames(res), ]res <- res/cellTypes_n_matrownames(res) <- paste(rownames(res), i, sep = "_")exprsMat_pseudo[[i]] = res}exprsMat_pseudo <- do.call(rbind, exprsMat_pseudo)return(exprsMat_pseudo)
}set.seed(1)
i =1
res1 <- create_pseudoBulk_parallel(exprsMat[, pseudobulk_sample ==pseudobulk_sample_list[i]], clust[[i]], k_fold = 30,use_bpparam = use_bpparam)set.seed(1)
i =1
res2 <- create_pseudoBulk_no(exprsMat[, pseudobulk_sample ==pseudobulk_sample_list[i]],clust[[i]], k_fold = 30)print("done")# for (i in seq_along(pseudobulk_sample_list)) {
#     res <- create_pseudoBulk_parallel(exprsMat[, pseudobulk_sample == 
#                                         pseudobulk_sample_list[i]], clust[[i]], k_fold = 30, 
#                              use_bpparam = use_bpparam)
# }

首先注意一个随机种子的问题，否则这个结果就会不一样