What expression spectrum is used in cox analysis?

Question

[1035655101]

Hi, all： I want to use TCGA's RNA-seq data for cox analysis, but the results calculated using FPKM are not the same as those calculated using TPM. I would like to ask which one is correct ？or， I am doing something wrong？

Answer 1

[Kevin Blighe]

It makes sense that you would obtain different results from FPKM and TPM. Neither of these are optimal for the purposes of statistical modelling.

I would encourage you to obtain the TCGA raw counts and then re-normalise using DESeq2. Then transform the normalised 'counts' via the regularised log transformation with blind=FALSE. It is on those regularised log expression levels that you would perform your Cox regression analysis.

Edit: the variance-stabilised expression levels would be fine, too.

Kevin