As Porf. Gordon suggested, CAMERA approach could handle the intercorrelation of genes, which will inflate p-values if GSEA. But this method only work with normal-limma voom data (pre-ranked GSEA within R? + Best DESeq2/limma-voom metric?).
(1) What is the "normal-limma voom data"?
(2) Besides, what is the shrink method used defalut in limma::voom?
(3) Does it also inflate the p value when using logFC conducted by limma::voom?
Any suggestions would be great appreciations!
camera() works for any sort of data that is analysable by limma or edgeR. cameraPR() works on any ranked list.
Your question seems to be based on overinterpretations of some 9-year old emails. In reality, voom and camera are not concerned with shrinkage, camera is not restricted to voom and there's so such thing as "voom data". While searching for previous posts is often a good idea, reading the camera documentation would be better! If you do read old emails, please be sure to concentrate on answers about voom rather than answers about other packages or emails that are actually questions.
Thanks for your reply sir! Is there any way to get the shrinkage logFC by limma like DESeq2?
We illustrate in the following workflows how to undertake GSEA using limma or edgeR:
These approaches rank genes by moderated t-test, which we prefer to shrunk logFC for this purpose.
Got that sir, thanks too much!