Seurat构建meta- feature表达矩阵

对于小数据集

直接使用aggregate函数对数据框形式的表达矩阵进行运算

counts <- sc[['RNA']]@counts #取出表达矩阵
4 x 4 sparse Matrix of class "dgCMatrix"
       cell.1 cell.2 cell.3 cell.4
gene.1      2      4      3      3
gene.2      6      4      3      3
gene.3      1      2      3      3
gene.4      8      9      .      3
data.use <- sc[['RNA']]@meta.features #取出需要重合的gene标签"pseudoid"
       pseudoid
gene.1        a
gene.2        a
gene.3        b
gene.4        c
 #根据pseudoid聚合基因表达量
counts.new = aggregate(counts,by=list(data.use$pseudoid),FUN=sum) %>% data.table::data.table() %>% tibble::column_to_rownames(var = "Group.1")
sc.agg <- CreateSeuratObject(counts = counts)

对于大型稀疏矩阵，推荐使用矩乘的形式进行处理

bMfeatures <- function(object,method="metaSum"){
    data <- object@assays$RNA@meta.features
    category.matrix <- Matrix::sparse.model.matrix(object = ~ -1 + data[,1] ) #创建分类矩阵
    category.matrix = Matrix::t(category.matrix)
    rownames(category.matrix) <- gsub("data\\[, 1\\]","",rownames(category.matrix))
    data.use <- GetAssayData(object, slot = "counts")
    aggreate.counts = category.matrix %*% data.use
    if (method == "gSI") {aggreate.counts = aggreate.counts / Matrix::rowSums(aggreate.counts)}#计算基因表达特异性指数
    toRet <- CreateSeuratObject(counts = aggreate.counts)
    return(toRet)
}
sc.meta <- bMfeatures(mini.sc,method = 'gSI')