Ancestry-Associated Differences in Somatic Mutation Rates From Tumor Profiling Data of A Pan-Cancer Cohort of 100,000 Patients

Authors Brooke Rhead, Yannick Pouliot, Justin Guinney, Francisco M. De La Vega

Racial and ethnic disparities span the continuum of cancer care and adversely impact screening, detection, diagnosis, treatment, and outcomes of cancer. These disparities are driven by a complex interplay among social, psychosocial, lifestyle, environmental, healthcare, and biological determinants of health. A challenge is that the social constructs of race and ethnicity may not fully capture shared genetic background. Genomics can capture ancestry in a more precise way, allowing genetic influences to be teased apart from the impact of social and environmental factors. Here, we searched for associations between continental genetic ancestry and somatic mutations in cancer genes in a pan-cancer cohort of 100,000 patients.

A sample of de-identified records from patients who underwent tumor profiling with the Tempus xT 648-gene assay was selected from the Tempus Database. We inferred global African (AFR), Amerindian (AMR), East Asian (EAS), European (EUR), and South Asian (SAS) continental ancestry proportions with a custom set of 658 ancestry informative markers. To properly account for the collinearity of ancestry proportions, we performed logistic compositional data analysis between mutation counts and ancestry proportions from data of cancer types with at least 300 patients and for genes for which a minimum of 10 and at least 1% of patients harbored variants. Odds ratios (ORs) represent the effect of a doubling in ancestry proportion, adjusted for age, gender, assay version, and all other ancestries. We examined associations for the presence of nonsynonymous somatic mutations and predicted driver mutations with the boostDM algorithm. We adjusted for multiple comparisons at a 5% false discovery rate (FDR) with the Benjamini-Hochberg method.

We identified previously reported significant (p<0.001) associations between EAS and predicted driver mutations in EGFR (OR=1.08) in lung cancer, and between AFR and APC (OR=1.03) and KRAS (OR=1.04) in colorectal cancer. Moreover, we found several novel associations, including AFR with TP53 (OR=1.06) and PTEN (OR=0.90) in endometrial cancer and BRAF (OR=0.93) in colorectal cancer, and EAS with CTNNB1 (OR=1.17) in lung cancer. The latter is noteworthy as CTNNB1, which has been proposed as a TTK inhibitor response biomarker, is overall infrequently mutated in lung cancer but activating mutations have been reported in East Asian patients. Indeed, we found that 7% of East Asian lung cancer patients in our cohort harbor a driver mutation in this gene vs. 1.6% in the other groups. Additionally, an association with nonsynonymous mutations was found between AMR with PTPN22 (OR=1.26) in colorectal cancer. While these associations can be the result of a combination of environmental, social, and biological factors, our results suggest that genetic ancestry can be useful to understand disparities in cancer incidence, progression, and outcomes.