Concordance Between Genomic Alterations Detected by Tumor and Germline Sequencing: Results from a Tertiary Care Academic Center Molecular Tumor Board

Abstract

Abstract Objective The majority of tumor sequencing currently performed on cancer patients does not include a matched normal control, and in cases where germline testing is performed, it is usually run independently of tumor testing. The rates of concordance between variants identified via germline and tumor testing in this context are poorly understood. We compared tumor and germline sequencing results in patients with breast, ovarian, pancreatic, and prostate cancer who were found to harbor alterations in genes associated with homologous recombination deficiency (HRD) and increased hereditary cancer risk. We then evaluated the potential for a computational somatic-germline-zygosity (SGZ) modeling algorithm to predict germline status based on tumor-only comprehensive genomic profiling (CGP) results. Methods A retrospective chart review was performed using an academic cancer center’s databases of somatic and germline sequencing tests, and concordance between tumor and germline results was assessed. SGZ modeling from tumor-only CGP was compared to germline results to assess this method’s accuracy in determining germline mutation status. Results A total of 115 patients with 146 total alterations were identified. Concordance rates between somatic and germline alterations ranged from 0% to 85.7% depending on the gene and variant classification. After correcting for differences in variant classification and filtering practices, SGZ modeling was found to have 97.2% sensitivity and 90.3% specificity for the prediction of somatic versus germline origin. Conclusions Mutations in HRD genes identified by tumor-only sequencing are frequently germline. Providers should be aware that technical differences related to assay design, variant filtering, and variant classification can contribute to discordance between tumor-only and germline sequencing test results. In addition, SGZ modeling had high predictive power to distinguish between mutations of somatic and germline origin without the need for a matched normal control, and could potentially be considered to inform clinical decision-making.