THE CANCER CELL MAP INITIATIVE v2.0
OVERALL SUMMARY
The Cancer Genome Atlas and sister projects have now sequenced over 20,000 tumor genomes, providing a
catalog of gene mutations, copy number variants and other genetic alterations associated with cancer. These
data have made it clear that every cancer is a distinct genetic disease, with tumors that look physiologically
similar often driven by patterns of gene mutations that are strikingly different. Due to this molecular heterogeneity,
it is typically unclear what are the key driver mutations or dependencies in a given cancer and how these
influence pathogenesis and response to therapy. One key observation for interpreting tumor genomes is that the
many rare tumor mutations can be shown to converge on common molecular networks. Based on this premise
we created the Cancer Cell Map Initiative (CCMI), whose mission is to create comprehensive maps of cancer
molecular networks and to use these maps in intelligent systems for personalized therapy. In 2017, the CCMI
was funded as an NCI U54 Research Center for Cancer Systems Biology, integrating expertise in network
mapping, bioinformatic analysis and cancer research from leading academic laboratories at two University of
California campuses (UCSF and UCSD). We have since generated comprehensive networks of protein
interactions in breast and head-and-neck tumor cells and, from these data, identified several hundred protein
complexes under selective mutational pressure in cancer (NeST v1.0). We have piloted deep learning systems
(DCell, DrugCell and TCRP) that can use this protein network information to translate a patient’s tumor mutation
profile to a predicted drug response, including FDA-approved and exploratory agents. We have implemented a
rich portfolio of training opportunities and, leveraging UC institutional support, expanded the CCMI consortium
to include more than a dozen faculty at UC and, most recently, Stanford. In the next five years, the CCMI will
seek to: (1) Generate comprehensive protein interaction networks centered on key cancer driver genes in lung
squamous cells (in healthy and diseased states) as well as the PIK3CA and TP53 pathways, which are central
to many tumor types; (2) Systematically extend the CCMI collection of cancer protein interaction data with protein
immunofluorescent imaging and cryo-electron microscopy to formulate multi-scale cancer cell maps; (3) Dissect
the functional logic of these networks and maps by systematic genetic screening experiments in the same tumor
types and pathways, using a panel of scalable cell proliferation, phenotype and pathway readouts; (4)
Significantly advance and harden our DrugCell interpretable deep learning system for cancer precision medicine;
(5) Train the current and next generation of scientists in network biology and its applications to cancer research;
and (6) Continue to build a cadre of leading investigators to expand CCMI into a global coordinated partnership.