I'm trying to do my own survival analysis with TCGA clinical data for PRAD in the following way:
- downloaded latest RNASeq V2 RSEM normalized for all samples
- calculated z-scores as z = (expression in tumor sample) - (mean expression in diploid samples) / (standard deviation of expression in diploid samples)
- downloaded latest clinical data
- followed "tutorial" here: http://bioinformatics.mdanderson.org/Supplements/ResidualDisease/Reports/osCurves.html
...however, using a threshold of +/- 1.96 for zscores, which should correspond to a p = 0.05, I have all the tumor samples that present at least one alteration in my gene signature...this means that I cannot really build a kaplain meier plot of altered vs non-altered samples...
how would you handle this situation?!