I would like to analyse whether the enrichment of a gene signature in TCGA data (bulk RNAseq data) is associated with survival.
To achieve this, I have performed ssGSEA to calculate enrichment scores for my gene signature within bulk RNAseq data (TCGA data), including 33 cancer types. To analyse whether the signature enrichment is associated with survival within the different cancer types, I performed Cox-regression separately for each of the 33 cancer types. I have read, however, that it is best to perform Z-score normalization of ssGSEA scores before survival analysis.
My question is now: Is it best to perform Z-score normalization before the Cox-regression across all cancer types, or should I perform Z-score normalization on a cancer-by-cancer-basis ?
Thank you very much for your help in advance!