The Oncogenic Potential and Selective Epistatic Interactions of Somatic Mutations can be Obtained from Human Tumor Sequence Data Alone

Somatic mutations are key contributors to tumorigenesis, driving tumor growth and progression. Quantifying how each somatic mutation impacts cancer progression is vital for the prioritization of targeted therapeutic interventions. Previous studies have argued that human cancer genomic data reveals the driver genes responsible for human cancer, but cannot reveal the fitness landscapes of tumorigenesis and underlying oncogene-tumor suppressor gene interactions. Based on this belief, autochthonous mouse models of lung adenocarcinoma (LUAD) have been used to quantify the tumorigenic potential of oncogenes, both alone and with dysfunctional tumor suppressor genes (TSGs). However, due to inherent genetic and physiological differences, and environmental factors, especially tobacco exposure, mouse models provide only a very coarse approximation of human biology. Therefore, it is worthwhile to rigorously evaluate what human cancer genomic data can reveal. To assess the oncogenic potential of oncogenes and TSGs in humans, we analyzed a comprehensive human LUAD dataset of 9424 samples, including 1082 smokers and 642 never-smokers. This analysis revealed substantial variations in the oncogenic contributions of KRAS G12C, KRAS G12D, BRAF V600E, and EGFR L858R. The effects of these mutations were influenced by other components of somatic genotype and by tissue exposure to a tobacco-smoke environment. KRAS G12D exhibited a higher effect size than KRAS G12C, consistent with findings from mouse models. EGFR L858R exhibited the highest effect sizes in all groups except smokers. BRAF V600E exhibited the lowest effect sizes except in never-smokers. Our analysis also uncovered selective epistatic interactions between oncogenic variants and TSGs, which were influenced both by oncogenic context and tobacco exposures. These findings emphasize the importance of considering patient-specific environmental factors in characterizing lung cancer's oncogenic landscape. In conclusion, human cancer genomic data can elucidate the tumorigenic potential of oncogenes and TSGs in humans–which is the species for which this information is most desirable. These findings highlight the capability of large human tumor sequence datasets to reveal complex interactions shaping cancer development and leverage human genetic and environmental diversity.