I am using DESeq2 for DE analysis in RNA-Seq experiment. I have 28 CLL patients raw counts and pooled male control and pooled female control raw counts. When I try to find differentially expressed miRNAs, I am getting wrong results (i.e. not matching with existing literature survey). e.g. hsa-mir-155 should be over-expressed but DESeq2 analysis shows under-expressed etc.
Here are my codes :
expDesign = data.frame(row.names = colnames(cts), condition = c("treated",..."treated","untreated","untreated")) dds = DESeqDataSetFromMatrix(countData = cts, colData = expDesign, design = ~condition) featureData = data.frame(gene=rownames(cts)) mcols(dds) = DataFrame(mcols(dds), featureData) dds = DESeq(dds) normalized = counts(dds, normalized=TRUE) res = results(dds)
When I run DESeq2 on my raw counts then I found following results:
LFC(mir-155) = -2.0649 LFC(mir-7e) = 3.5927 LFC(mir-30a) = 4.1988
But after shrinkin LFC, I got the following result:
LFC(mir-155) = 12.3063 LFC(mir-7e) = 5.2165 LFC(mir-30a) = 6.9991
My questions are as follows:
- Which one of above LFC is I should consider as correct
- How to decide whether I should go for shrunken LFC
- How to decide the experiment design as group and condition. Here I have two conditions only (either treated or untreated).
- What is the role of normal while doing expression analysis in DESeq2.
Guidance much needed. Thanks.