Geneset-based cancer survival analysis: On the non-uniform distributiuon of p-values under the null hypothesis

Although null p-values are assumed to be uniformly distributed in gene expression experiments, the actual distribution often deviates from the assumed distribution. This can incorrectly associate the biology of a geneset with cancer prognosis in geneset-based survival studies. To assess the implications of this, a geneset-based method was developed. This method empirically approximates the distribution of null p-values and tests whether predefined sets of biologically-related genes are associated with survival.