BackgroundIdentification of tumor heterogeneity and genomic similarities across different cancer types is essential to the design of effective stratified treatments and for the discovery of treatments that can be extended to different types of tumors. However, systematic investigations on comprehensive molecular profiles have not been fully explored to achieve this goal.ResultsHere, we performed a network-based integrative pan-cancer genomic analysis on >3000 samples from 12 cancer types to uncover novel stratifications among tumors. Our study not only revealed recurrently reported cross-cancer similarities, but also identified novel ones. The macro-scale stratification demonstrates strong clinical relevance and reveals consistent risk tendency among cancer types. The micro-scale stratification shows essential pan-cancer heterogeneity with subgroup-specific gene network characteristics and biological functions.ConclusionsIn summary, our comprehensive network-based pan-cancer stratification provides valuable information about inter- and intra- cancer stratification for patient clinical assessments and therapeutic strategies.Electronic supplementary materialThe online version of this article (doi:10.1186/s12864-015-1687-x) contains supplementary material, which is available to authorized users.