Ancestry-adjusted genomic analysis to identify known and novel associations with single and multiple cancer phenotypes.

Abstract

10504 Background: Germline cancer risks are impacted by cancer type, stage, specific mutation, and ancestry. This study aims to determine the magnitude of genetic risks for commonly tested genes across patients with single and multiple primary cancers (MPC). Methods: We conducted a partially de-identified, retrospective cohort study of all patients consented to using the MSK-IMPACT, a matched tumor-normal sequencing assay, over a five-year period. Approximately 83% of IMPACT testing is derived from patients with metastatic disease. We collected data on demographics and clinical cancer features through cBioPortal and institutional databases. MPC were classified based on the IARC guidelines. Clinically actionable, pathogenic variants in cases and gnomAD non-cancer controls were identified using PathoMAN. Statistical methods incorporated ancestry adjusted, weighted generalized linear model, and ancestry specific methods to estimate risk. Biallelic losses were analyzed from paired tumors. Results: We analyzed 28,000 individuals across 35 cancer types, where germline variants were discovered within 100 cancer predisposition genes. 16% carried pathogenic, clinically actionable variants in known and novel cancer predisposition genes (15% in those with single primary and 21% in MPC). We discovered statistically significant novel associations of germline variants with several cancer phenotypes, including BRCA2 and biliary cancer (OR = 4.5, CI = 2.2-8); ATM and BRCA2 with mesothelioma (OR = 4.8, CI = 2.1-9.2; OR = 8, CI = 3.1-16.6, respectively); NF1 and breast cancer (OR = 5.8, CI = 2.4-12.1); NBN and lung cancer (OR = 4.2, CI = 2.3-6.9), bladder with ATR and others. We replicated known associations and discovered pleiotropy of BRCA1/2, NF1, ATM, TP53 with several cancer types. We failed to replicate reported associations such as PALB2 and ovarian cancer and CHEK2 and colon cancer. Amongst cancer pairs of multiple primaries, we observed mismatch repair genes MSH2, MSH6 and MLH1 to be associated with increased risks for stomach, bowel, and uterine cancer pairs, and TP53 in breast MPC compared to breast single primaries. Conclusions: An ancestry adjusted, weighted analysis of a large real-world case-cohort enriched with metastatic disease did not confirm some prior genotype-phenotype correlations, while suggesting novel associations with hereditary cancer types.