Abstract 1176: Leveraging the GENIE dataset to distinguish somatic cancer drivers from passenger events in routine oncology practice

• Published in: Cancer research (Chicago, Ill.) Ill.), 2022-06, Vol.82 (12_Supplement), p.1176-1176
• Main Authors: Beer, Philip A., Cooke, Susanna L., Li, Xuan Shirley, Biankin, Andrew V.
• Format: Article
• Language: English
• ISSN: 1538-7445; 1538-7445
• DOI: 10.1158/1538-7445.AM2022-1176

Distinguishing genomic events that drive cancer (drivers) from inconsequential damage (passengers) is key to matching patients with biomarker-associated therapies and clinical trials. Whilst population level genomic analyses can estimate the proportion of driver events within a given gene, they cannot categorize individual variants. Reliable annotation remains a significant barrier to unlocking the full utility of cancer genomics. GENIE (Genomics Evidence Neoplasia Information Exchange) data was used to generate codon recurrence scores (CRS) for known cancer genes. GENIE was parsed to remove duplicates, sequencing artefact (recurrent variants reported by one institution, invariably associated with amplicon-based sequencing) and hypermutated samples (>15 mutations per megabase). For missense mutations, a codon recurrence of ≥10 was used to classify driver mutations. The table shows proportion of missense mutations classified as drivers by gene and cancer type, comparing population-based non-synonymous to synonymous mutation ratio; dN/dS (Martincorena 2017, PMID 29056346) with assessment of individual variants by CRS. This comparison demonstrates a high degree of concordance. For a few genes, CRS under called driver mutations (ie VHL in renal cancer) likely due to reduced power with small sample numbers. Importantly, CRS is able to assign driver status to individual missense mutations. Comparison with informatic analyses (PMIDs 28115009, 29247016, 30365005, 31034466) showed equivalent or superior performance of the GENIE approach. These findings demonstrate how a large (and expanding) real-world dataset can be used to predict the driver status of somatic missense mutations at the n=1 variant level. This process is amenable to implementation as a rules-based classification process for somatic missense mutations as part of an automated annotation pipeline. +corrected dN/dS ratio not significantly different from one. Pan- cancer Breast cancer Renal cancer Lung adenocarcinoma Gene dN/dS CRS dN/dS CRS dN/dS CRS dN/dS CRS BRAF 91% 76%