I have a TCGA raw RNA-seq counts (HTseq) for tumor samples. I want to do survival analysis based on gene expression for a particular gene set.
I know that I can download normalized TCGA data with z-scores etc. But as I have a certain set of genes I would like to separate them based on the median expression of a gene - into "UP" (greater than the median) and "DOWN" (lower than the median).
In all similar tutorials or answers here, I see that people do differential analysis using DESeq or EdgeR and use built-in normalization. I do not need to do differential analysis, I just need to find associations between particular genes' expression and overall survival.
Can I use just some simple data normalization like CPM or log normalization on raw RNA-seq counts to do survival analysis as I described above?
0 answers
No answers yet.
Log in to answer this question.
Hello,
Did you find an answer for this? Ideally any manual/solid giving solid guidelines? I am facing the same problem.
Thank you!
It's just edgeR or DESeq2 normalization. If you want to compare medians or any other metric then data have to be corrected for depth and composition -- exactly what mentioned tools do. Use the logcpms from edgeR or vst from DESeq2 and then do your median/quartile/whatever thingy on it. Don't overthink it.