RNA-Seq data from Firebrowse TCGA

Question

[landscape95]

Hi, I am building the logistic regression between radiotherapy clinical data (yes/no) and gene expression for the probability of assigning RT to patients from RNA-Seq from Firebrowse TCGA Breast cancer RSEM normalized, but I think it the values of the genes vary a lot (please have a look in the picture). Should I do any normalization to balance between these gene expression? Thanks a lot!

Answer 1

[Kevin Blighe]

I would import these RSEM gene-level [estimated] counts to DESeq2 either manually or via tximport. This is mentioned in the vignette. In DESeq2, I would normalise them with RT in my design formula and then transform these normalised counts to regularised log expression levels with blind = FALSE, i.e., rlog(dds, blind = FALSE).

It is those transformed, normalised expression levels on which I would conduct my downstream regression analysis.

Kevin