单细胞转录组/FindMarkers

FindMarkers

gzh：BBio

对分群结果进行差异基因鉴定的函数，理想情况下，对于每群细胞来说，marker基因都位于上调基因的前列。

#FindMarker结果以p_val从小到大排列，wilcox检验方法使用data中的数据。pct.1和pct.2则分别为基因在对应ident参数中的表达比例。
df <- FindMarkers(pbmc, ident.1=1, ident.2=0, slot="data", logfc.threshold=0.25, min.pct=0.1, test.use = "wilcox")
head(df)
# p_val avg_logFC pct.1 pct.2    p_val_adj
# TYMP     1.702818e-11  2.539289  1.00 0.111 3.916481e-09
# CST3     4.469249e-11  2.552769  1.00 0.306 1.027927e-08
# S100A8   5.334985e-11  4.037048  0.96 0.111 1.227047e-08
# LYZ      6.997602e-11  3.082150  1.00 0.417 1.609449e-08
# HLA-DRB1 3.287672e-10  3.325130  0.88 0.083 7.561646e-08
# HLA-DPB1 4.061018e-10  3.547416  0.88 0.083 9.340340e-08

avg_logFC

结果中avg_logFC是否是FC的自然对数呢？看看源码吧。

getAnywhere('FindMarkers.default')
# mean.fxn <- if (is.null(x = reduction) && slot != "scale.data") {
#         switch(EXPR = slot, data = function(x) {
#             return(log(x = rowMeans(x = expm1(x = x)) + pseudocount.use))
#         }, function(x) {
#             return(log(x = rowMeans(x = x) + pseudocount.use))
#         })
#     }
#     else {
#         rowMeans
#     }
#     data.1 <- mean.fxn(data[features, cells.1, drop = FALSE])
#     data.2 <- mean.fxn(data[features, cells.2, drop = FALSE])
#     total.diff <- (data.1 - data.2)

logfc.threshold和min.pct

差异倍数及表达比例的参数设置会影响运行时间吗？

system.time(FindMarkers(pbmc, ident.1=1, logfc.threshold=0.25, min.pct=0.1))
#user  system elapsed 
#21.154   0.435  21.590
system.time(FindMarkers(pbmc, ident.1=1, logfc.threshold=0, min.pct=0))
#user  system elapsed 
#330.599   1.365 332.409

放宽参数的设置耗时大幅增加，查看源码。min.pct和logfc.threshold都在代码头部用于过滤基因，阈值放宽使得用于分析的基因增多。

马克marker

#B细胞FeaturePlot(object = pbmc_small, features = c('MS4A1', 'CD19', 'CD79B'),ncol=3)