Abstract 2182: Ancestry inference and population-specific disparities in a real-world clinical sequencing cohort

Accurate ancestry inference is crucial for identifying genetic determinants of cancer disparities among specific populations. While methods to infer genetic ancestry and admixture have historically relied upon genome-wide markers from broad-scale next-generation sequencing (NGS), the adaptation to targeted NGS panels presents an opportunity to prospectively incorporate ancestry inference as part of routine clinical diagnosis. Here we show that global ancestral contributions and admixture of African (AFR), European (EUR), East Asian (EAS), Native American (NAM) and South Asian (SAS) populations can be reliably inferred using markers from genomic regions covered by the FDA-authorized clinical NGS panel, MSK-IMPACT. We also show that individuals with Ashkenazi Jewish (ASJ) ancestry can be inferred with 97% accuracy using a set of ASJ ancestry-informative markers. We apply these methods to infer genetic ancestry for over 45,000 patients from the AACR GENIE v10.0-public cohort who were profiled using MSK-IMPACT. We observed 98% concordance between self-reported race and genetic ancestry for non-admixed individuals and were able to quantitatively infer ancestral contributions for individuals from recently admixed populations such as African American and Latin American. Of the discordant cases, manual review of clinical and family histories revealed the vast majority to represent clinical encoding errors where the inferred ancestry was confirmed correct. As self-reported race is not available for 17% of patients in the GENIE cohort, the ability to accurately infer genetic ancestry enables broader analysis of population-specific genomic and clinical features. We systematically evaluated the frequency of somatic and germline alterations in up to 468 genes for each cancer type and recapitulated known differences among ancestral populations. For example, we observed significantly higher frequency of EGFR somatic alterations in EAS (65%) and SAS (66%) compared to non-ASJ EUR (21%) with lung adenocarcinoma, a difference that remained significant even when limiting to never-smokers. Additionally, we found that the frequency of harboring at least one known BRCA founder mutation (BRCA1 68_69delAG, BRCA1 5266dupC, or BRCA2 5946delT) was 26x higher in genetically determined ASJ (5.1%) compared to non-ASJ (0.2%). Strikingly, while the overall rate of driver alterations in solid tumors was similar across different populations, we found that the proportion of patients with cli