Interpretation of cancer mutations using a multiscale map of protein systems

Abstract

Description Mapping protein interactions driving cancer Cancer is a genetic disease, and much cancer research is focused on identifying carcinogenic mutations and determining how they relate to disease progression. Three papers demonstrate how mutations are processed through networks of protein interactions to promote cancer (see the Perspective by Cheng and Jackson). Swaney et al. focus on head and neck cancer and identify cancer-enriched interactions, demonstrating how point mutant–dependent interactions of PIK3CA, a kinase frequently mutated in human cancers, are predictive of drug response. Kim et al. focus on breast cancer and identify two proteins functionally connected to the tumor-suppressor gene BRCA1 and two proteins that regulate PIK3CA. Zheng et al. developed a statistical model that identifies protein networks that are under mutation pressure across different cancer types, including a complex bringing together PIK3CA with actomyosin proteins. These papers provide a resource that will be helpful in interpreting cancer genomic data. —VV Large-scale protein interaction maps provide a framework for interpreting cancer genomic data. INTRODUCTION Tumor genome sequencing has revealed that, beyond a few commonly mutated genes, most mutations that affect cancer genomes are rare. To interpret these rare events, a powerful approach has been to organize mutations by their effects on commonly dysregulated cellular systems. Understanding the cancer genome in this way requires surmounting two challenges: (i) How do we comprehensively map cancer cell systems? (ii) How do we identify which systems are under mutational selection? RATIONALE To address these questions, we used proteomic mass spectrometry and data integration to build a structured map of protein assemblies found in human cancer cells. We then developed a statistical model of mutation, pinpointing which assemblies are under strong mutational selection and in which cancer types. The goal was to interpret the many rare gene mutations that affect tumor genomes by their convergence on higher-order entities. RESULTS We amassed a large compendium of cancer protein interactions, combining the screens in breast cancer (Kim et al., this issue) and head-and-neck cancer (Swaney et al., this issue) with multi-omic evidence from 127 previous studies. Lines of evidence were integrated quantitatively to yield a continuous metric of association for each protein pair (integrated association stringency, or IAS). This network of protein associations exhibited clear multiscale and modular structure, revealing 2338 robust assemblies of interacting proteins (hereafter “protein systems”) across different stringencies. Systems were organized hierarchically, with small high-stringency systems (e.g., specific complexes) combining in larger ones (e.g., processes and organelles) as stringency was relaxed. We next developed a statistical model, HiSig, to identify a parsimonious set of systems that best explains the gene mutation frequencies observed in tumors. HiSig analysis of 13 tumor types yielded a map of 395 mutated protein systems we call NeST (Nested Systems in Tumors, http://ccmi.org/nest/). NeST comprised numerous small complexes, most mutated within specific tumor types, organized within larger systems relevant to most cancers. Although NeST recapitulated cancer hallmarks, the majority of systems had not been previously described or had not been associated with cancer mutation. Nonetheless, many were recurrently mutated in independent cohorts, supporting their significance. Notable systems included a PIK3CA-actomyosin complex that points to a new mode of phosphatidylinositol 3-kinase regulation, as well as recurrent mutations in collagen complexes that we found to disrupt the extracellular matrix, thereby promoting proliferation. Finally, we identified NeST systems that serve as biomarkers of cancer outcomes, leading to 548 genes for potential use in clinical sequencing panels. CONCLUSION In their classic description of the “Hallmarks of Cancer,” Hanahan and Weinberg predicted that the “complexities of cancer ... will become understandable in terms of a small number of underlying principles.” Around the same time, Alberts provided his seminal perspective of the cell as a collection of “protein assemblies [interacting] in an elaborate network.” By organizing disparate tumor mutations into underlying principles captured by a multiscale map of protein assemblies, this work represents a synthesis of these visions. The strategies developed here may generalize to other diseases that are affected by rare genetic alterations. Mapping cancer protein systems. Protein interaction datasets were integrated to identify protein communities (“systems”) at multiple scales of analysis (left). Each system was tested for cancer mutational selection as a system versus its substituent proteins, revealing a hierarchy of protein systems under selection in cancer (center). Discoveries from this hierarchy (right) were validated with clinical data and functional experiments. A major goal of cancer research is to understand how mutations distributed across diverse genes affect common cellular systems, including multiprotein complexes and assemblies. Two challenges—how to comprehensively map such systems and how to identify which are under mutational selection—have hindered this understanding. Accordingly, we created a comprehensive map of cancer protein systems integrating both new and published multi-omic interaction data at multiple scales of analysis. We then developed a unified statistical model that pinpoints 395 specific systems under mutational selection across 13 cancer types. This map, called NeST (Nested Systems in Tumors), incorporates canonical processes and notable discoveries, including a PIK3CA-actomyosin complex that inhibits phosphatidylinositol 3-kinase signaling and recurrent mutations in collagen complexes that promote tumor proliferation. These systems can be used as clinical biomarkers and implicate a total of 548 genes in cancer evolution and progression. This work shows how disparate tumor mutations converge on protein assemblies at different scales.