谢谢Michael的回复。

虽然我已经看到了使用duplicateCorrelation推荐limma的讨论，但我想知道我是否可以用DESeq2来做。

我的困惑恰恰是在小插图“群体特定条件的影响，群体中嵌套的个体”的部分，因为我认为主体。n是将两个不同的科目分组，但我现在已经意识到它到底是做什么的，所以没关系。但我并不是在寻找特定于群体(在我的情况下是特定于批次)的条件效应，而是在调整群体/批次和受试者差异时的条件效应。所以我在添加主题后尝试了下面的模型。n变量:

~ batch + batch:主语。N +条件

但让我再往前走一步，在简单模型中，我遇到了一个收敛问题，为此我尝试了两种方法，但都没有解决问题。首先，将DESeq函数分解为步骤并增加maxit，其次，过滤更多的低表达基因。

#对象构造的简化模型:dds = DESeqDataSetFromMatrix(countData = counts. frommatrix)mat, colData = sample.summary。平衡，设计= ~受试者+条件)dds <- DESeq(dds) *估计大小因子估计分散基因方面的色散估计均值-色散关系最终的色散估计拟合模型和测试21行不收敛于beta，标记为mcols(对象)$betaConv。使用更大的maxit参数与nbinomWaldTest* #所以我重复分成步骤和过滤更多的低表达基因:dds = DESeqDataSetFromMatrix(countData =计数。mat, colData = sample.summary。平衡，设计= ~受试者+条件)dds <- estimateSizeFactors(dds) nc <- counts(dds，归一化=TRUE) filter <- rowsum (nc >= 10) >= 4 dds <- dds[filter，] dds <- estimatedispersion (dds) *基因色散估计均值-色散关系最终色散估计* dds <- nbinomWaldTest(dds, maxit=2000) *23行未在beta中收敛，标记在mcols(object)$betaConv。使用更大的maxit参数nbinomWaldTest* #没有解决收敛问题，我尝试继续前进。#添加组合变量dds ColData: dds$subject。N = ifelse(dds$batch == 1, factor(rep(rep(1:batch1.subjects,each=2)，2))， factor(rep(rep(1:batch2.subjects,each=2)，2)))N = as.factor(dds$subject. N)样本批处理条件n 1 1 sample1 S1 2 sample2 1 B S1 1 3 sample3 1 S2 2 4 sample4 1 B S2 2 5 1 sample5 S3 3 1 6 sample6 1 B S3 3 7 sample7 S4 4 8 sample8 1 B S4 4 9 sample9 1 S5 5 10 sample10 1 B S5 5 11 sample11 2 S6 1 12 sample12 2 B S6 1 13 sample13 2 S7 2 14 sample14 2 B S7 2 #饲料新colData DESeqDataSet和编辑设计:dds = DESeqDataSetFromMatrix (countData =。mat, colData = colData(dds)， design = ~ batch + batch*subject。N +条件) *converting counts to integer mode Error in checkFullRank(modelMatrix) : the model matrix is not full rank, so the model cannot be fit as specified. One or more variables or interaction terms in the design formula are linear combinations of the others and must be removed.* # This is because I have unbalanced data in respect to batch: uneven subjects in the two batches. # So I removed the levels without samples, like for example phase2:subject.n3: m = model.matrix(~batch + batch:subject.n + condition, colData(dds)) all.zero <- apply(m, 2, function(x) all(x==0)) length(all.zero[all.zero == TRUE]) idx <- which(all.zero) m <- m[,-idx] # Feed clean model matrix to DESeq: dds = DESeq(dds, full = m) *using supplied model matrix using pre-existing size factors estimating dispersions gene-wise dispersion estimates mean-dispersion relationship final dispersion estimates fitting model and testing 121 rows did not converge in beta, labelled in mcols(object)$betaConv. Use larger maxit argument with nbinomWaldTest* resultsNames(dds) [1] "Intercept" "batch2" "conditionB" [4] "batch1.subject.n2" "batch2.subject.n2" "batch1.subject.n3" [7] "batch1.subject.n4" "batch1.subject.n5"

如何解决收敛问题?我甚至重复了单独的步骤，并添加了useOptim选项:

dds <- estimateSizeFactors(dds) dds <- estimateDispersions(dds) dds <- nbinomWaldTest(dds, maxit=2000, useOptim = T)

另外，为了校正批次和受试者的差异，是res1还是res2?

res2 = results(dds, contrast= c("condition"， "B"， "A"))n2”、“batch2.subject。n2”、“batch1.subject。n3”、“batch1.subject。陶瓷”、“batch1.subject。N5 "， "batch2")))