Summary
Germline variants have been recognized as a key determinant in cancer susceptibilities, patient prognosis and treatment outcome. Previous studies have highlighted significant differences in pathogenic variant frequencies across different populations, underscoring the influence of ancestry-specific genetics on cancer risk, survival and drug response. Leveraging extensive real-world data from ConcertAI® electronic health records (EHR) linked with Caris Life Sciences genomic data, we analyzed germline variants in a diverse multi-cancer cohort of 2,776 tumors, including five indications: non-small cell lung cancer (NSCLC, 1,514), colorectal cancer (860), pancreatic (218), gastroesophageal (96), and breast (88) cancers. Notably, we observed a higher proportion of African Americans in breast cancer and more males with gastroesophageal and pancreatic cancers, mirroring US cancer incidence across racial and gender groups.
Following germline variant calling by DNA-seq pipeline and annotation by snpeff and snpsift dbSNP, we filtered out low quality and suspected somatic mutations. Utilizing the ClinVar database and the CharGer tool, we identified pathogenic variants in cancer risk genes for each cancer type. When comparing the frequencies of pathogenic variants in ASCO-recommended high-risk genes between our pan-cancer cohort and TCGA, we found that the frequencies were generally comparable. Subsequently, we tested for the association of these pathogenic variants with patient overall survival. Statistically significant associations were detected between patient poor survival and pathogenic variants in STK11 (p = 0.005, in lung NSCLC), SMAD4 (p = 0.013, in colorectal cancer), and CDKN2A (p = 0.034, in pancreatic cancer), alongside novel associations with PTH (p = 0.012 and 0.024 in gastroesophageal and pancreatic cancers) and HFE (p = 0.037, in pancreatic cancer). Intriguingly, Asian patients exhibited a much higher frequency of PTH variants (allele frequency > 0.30 %) than other races, while HFE variants were more common in Hispanic/Latino populations, reported to cause hereditary hemochromatosis. Although no significant association with immunotherapy (IO) response was detected in the curated NSCLC IO response cohort, the potential links with additional IO markers (e.g., tumor mutational burden, immunologic constant of rejection score, Miracle score) warrant further exploration.
Our study describes an innovative approach for investigating the relationships between germline variants and cancer patient prognosis and treatment responses using the EHR linked genomics data — an area often under-explored. Our approach integrates genetic profiling with comprehensive clinical data to characterize germline variants that may serve as predictive biomarkers, ultimately aiming to advance personalized medicine and improve patient outcomes.