如何克服 shrink , in deseq ？

Question

目前正在尝试两组间的差异表达。此代码在 6 个月前有效，但现在我得到： deseq2 库

> converting counts to integer mode estimating size factors estimating
> dispersions gene-wise dispersion estimates mean-dispersion
> relationship final dispersion estimates fitting model and testing
> -- replacing outliers and refitting for 7129 genes
> -- DESeq argument 'minReplicatesForReplace' = 7 
> -- original counts are preserved in counts(dds) estimating dispersions fitting model and testing
> **Error in lfcShrink(dds = this.dds, contrast = c("condition", G1.desc,  :    type='apeglm' shrinkage only for use with 'coef' In
> addition: Warning message: In DESeqDataSet(se, design = design,
> ignoreRank) :   some variables in design formula are characters,
> converting to factors**



 this.dds <- DESeqDataSetFromMatrix(countData = this.dat,
                                         colData = this.coldata,
                                         design = ~ condition)
      
      this.dds$condition <- factor(this.dds$condition, levels = c(G1.desc,G2.desc))
      
      if (! exists("Dseq_results")) {
        Dseq_results <- DESeq(this.dds)
      }
      this.dds <- Dseq_results
      res <- results(this.dds, 
                     name=sprintf("condition_%s_vs_%s",G1.desc,G2.desc),contrast = c('condition',G1.desc,G2.desc)
      )
      
      tmp1 <<- this.dds
      print(tmp1)
      res1 <- lfcShrink(dds=this.dds,contrast = c('condition',G1.desc,G2.desc), res=res)
  
  
  res1.nice=cbind(rownames(HTseq.counts),res1$baseMean,res1$log2FoldChange,res1$lfcSE,res1$stat,res1$pvalue,res1$padj)
  colnames(res1.nice)<-c('Gene','baseMean','log2FoldChange','lfcSE','stat','pvalue','padj')
  
  return(res1.nice)

g1,g2 是组，desc - 描述，this.dds - 保存 htseq 计数的数据。我尝试将类型中的类型从“apeglm”更改为 'ashr'，但出现了不同的错误。没有打印任何内容。

Answer 1

举个例子，我真的不确定你对 G1.desc 有什么要求，但可以说：

G1.desc = "A"
G2.desc = "B"

this.dat = matrix(rnbinom(3000,mu=50,size=0.1),ncol=6)
colnames(this.dat) = paste0("s",1:6)
this.coldata = data.frame(condition = rep(c("A","B"),each=3))

在你把它放入deseq对象之前的因素，在这里你已经可以决定对比度了：

this.coldata$condition = factor(this.coldata$condition,
levels=c(G1.desc,G2.desc))

运行 DESeq:

this.dds <- DESeqDataSetFromMatrix(countData = this.dat,
                                   colData = this.coldata,
                                   design = ~ condition)
      
this.dds <- DESeq(this.dds)
res <- results(this.dds, 
name=sprintf("condition_%s_vs_%s",G1.desc,G2.desc),
contrast = c('condition',G1.desc,G2.desc)
      )

是的，我在“apeglm”上遇到了与您相同的错误：

l2fc = sapply(c("normal","ashr","apeglm"),function(i){
lfcShrink(dds=this.dds,
contrast = c('condition',G1.desc,G2.desc), res=res,type=i)$log2FoldChange
})

Error in lfcShrink(dds = this.dds, contrast = c("condition", G1.desc,  : 
  type='apeglm' shrinkage only for use with 'coef'

如果你像我上面那样设置因素，对比度总是G2.desc vs G1.desc，你可以看到：

resultsNames(this.dds)
[1] "Intercept"        "condition_B_vs_A"

所以你只是不提供 res = 参数：

 l2fc = sapply(c("normal","ashr","apeglm"),function(i){
    lfcShrink(dds=this.dds,coef=2,type=i)$log2FoldChange
    })

head(l2fc)
         normal       ashr      apeglm
[1,] -0.2313623 -0.1739082 -0.02573008
[2,] -1.0543602 -0.8159166 -0.11381314
[3,]  2.2493688  1.8142392  0.24608219
[4,]  3.5059792  3.0354023  0.22795819
[5,]  2.8617158  2.3939858  0.32407906
[6,] -0.2236092 -0.1710536 -0.02053698

如何克服 shrink , in deseq ？

How to overcome shrink , in deseq?

r

coefficients