padj value in DEseq package

Question

[Varun Gupta]

Hi Everyone,

I am using DEseq package for my RNA SEQ samples so as to find differentially expressed genes between 2 samples

I have 2 RNA seq samples. One of them is CONTROL and the other sample is prepared by UV irradiation. Lets call the other sample as X.

Now since I don't have biological replicates in my case, I will have to use DEseq PACKAGE (working without any replicates section).

I randomly gave 2 conditions, "control" and "x" to my sample and I am looking for differentially expressed genes between 2 samples(up regulated or down regulated).

After following all the steps of DEseq at the last step where final results are to be computed

res <- nbinomTest(data,"control","x")

In the final output padj is always 1

padj - p-value adjusted for multiple testing using Benjamini-Hochberg to estimate the false discovery rate

Should I take into account only the pval column in my case or is there any other way to analyze this situation where I want to find differentially expressed genes between 2 samples(without biological replicates).

Regards
Varun

Answer 1

[bw.]

Its difficult to know statistical significance (eg. p-value) of differences in expression when you don't have replicates.

One thing you can do to maybe get some meaning out of the data is sort your genes by the unadjusted p-value, or by fold-change, and look at the top/bottom genes. You can also feed this sorted list to GSEA to check for gene-set enrichment.

[Varun Gupta]

Well that is what I am doing now to analyze top/bottom genes by fold-change. Also are you talking about GSEA R PACKAGE called as GSEABase?? Tell me about it and I will try it. Thanks

[bw.]

Its gene set enrichment analysis (GSEA) from the Broad Inst. GSEABase is the R interface for it.

All this depends on what questions you want to answer based on this data...