Purpose: The Cancer Genome Atlas data resources represent an opportunity to explore commonalities across cancer types involving multiple molecular levels, but tumor lineage and histology can represent a barrier in moving beyond differences related to cancer type.

Experimental Design: On the basis of gene expression data, we classified 10,224 cancers, representing 32 major types, into 10 molecular-based “classes.” Molecular patterns representing tissue or histologic dominant effects were first removed computationally, with the resulting classes representing emergent themes across tumor lineages.

Results: Key differences involving mRNAs, miRNAs, proteins, and DNA methylation underscored the pan-cancer classes. One class expressing neuroendocrine and cancer-testis antigen markers represented ∼4% of cancers surveyed. Basal-like breast cancers segregated into an exclusive class, distinct from all other cancers. Immune checkpoint pathway markers and molecular signatures of immune infiltrates were most strongly manifested within a class representing ∼13% of cancers. Pathway-level differences involving hypoxia, NRF2-ARE, Wnt, and Notch were manifested in two additional classes enriched for mesenchymal markers and miR200 silencing.

Conclusions: All pan-cancer molecular classes uncovered here, with the important exception of the basal-like breast cancer class, involve a wide range of cancer types and would facilitate understanding the molecular underpinnings of cancers beyond tissue-oriented domains. Numerous biological processes associated with cancer in the laboratory setting were found here to be coordinately manifested across large subsets of human cancers. The number of cancers manifesting features of neuroendocrine tumors may be much higher than previously thought, which disease is known to occur in many different tissues. Clin Cancer Res; 24(9); 2182–93. ©2018 AACR.