Hi,

I have run DE analysis using different input files. One is with multiple conditions and another one is just pairwise comparison between treatment and control. I found that the p-value is kind of different. The input files are as below:

For multiple conditions, the input file is as below:

condition       replicate       batch
Ctrl_A  1       1
T1_A    1       1
T2_A    1       1
Ctrl_B  1       1
T1_B    1       1
T2_B    1       1
Ctrl_C  1       1
T1_C    1       1
T2_C    1       1
Ctrl_D  1       1
T1_D    1       1
T2_D    1       1
Ctrl_A  2       2
T1_A    2       2
T2_A    2       2
Ctrl_B  2       2
T1_B    2       2
T2_B    2       2
Ctrl_C  2       2
T1_C    2       2
T2_C    2       2
Ctrl_D  2       2
T1_D    2       2
T2_D    2       2
Ctrl_A  3       3
T1_A    3       3
T2_A    3       3
Ctrl_B  3       3
T1_B    3       3
T2_B    3       3
Ctrl_C  3       3
T1_C    3       3
T2_C    3       3
Ctrl_D  3       3
T1_D    3       3
T2_D    3       3

I would like to compare only T1_D and Ctrl_D. The code is as below:

directory="./"
datList=read.table("multiple.input",header=TRUE)
sampleTable=data.frame(datList)

ddsHTSeq <- DESeqDataSetFromHTSeqCount(sampleTable = sampleTable,
                                       directory = directory,
                                       design= ~ batch + condition)

ddsHTSeq$condition <- factor(ddsHTSeq$condition)
#dds <- estimateSizeFactors(ddsHTSeq)
#nc <- counts(dds, normalized=TRUE)
#filter <- rowSums(nc >= 1) >= 1
#dds <- dds[filter,]
dds <- DESeq(ddsHTSeq)
res <- results(dds, contrast=c("condition", "T1_D","Ctrl_D")
resOrdered <- res[order(res$pvalue),]
resSig <- subset(resOrdered, padj < 0.1)

For pairwise conditions (T1_D vs. Ctrl_D), the input file is as below:

condition       replicate       batch
Ctrl_D  1       1
T1_D    1       1
Ctrl_D  2       2
T1_D    2       2
Ctrl_D  3       3
T1_D    3       3

The code is as below:

directory="./"
datList=read.table("pairwise.input",header=TRUE)
sampleTable=data.frame(datList)

ddsHTSeq <- DESeqDataSetFromHTSeqCount(sampleTable = sampleTable,
                                       directory = directory,
                                       design= ~ batch + condition)

ddsHTSeq$condition <- relevel(ddsHTSeq$condition, ref = "Ctrl_D")
dds <- DESeq(ddsHTSeq)
res <- results(dds)
resOrdered <- res[order(res$pvalue),]
resSig <- subset(resOrdered, padj < 0.1)

I found that the number of genes with significant expression is different due to difference in p-value and adjusted p-value when the multiple conditions and pairwise conditions are used separately. Is there some error in the code? Thanks a lot.

This is expected and it’s  one of the FAQ in the vignette.