Advancing precision medicine in the treatment of cancer is complex and requires extensive molecular profiling to identify therapeutic targets and appropriately stratify patients into groups that differ in therapeutic vulnerability. The Cancer Genome Atlas (TCGA) has molecularly defined several cancer types and identified distinct subtypes. However, in some cases identification of molecular subtypes have failed to translate clinically. We hypothesize that more accurate molecular subtyping and patient stratification can be achieved by taking into account intra-tumoral heterogeneity and microenvironment signals.