Enhancing variant interpretation through multi-database and systematic variant classification: Reducing uncertainty in clinical genomics

Background Interpretation of Variants of unknown significance (VUS) is one of the major challenge in the era of comprehensive genomic profiling (CGP) in precision oncology. With several broad genomic panels hitting the oncology market, most struggle with VUS and its clinical utility. Thus, it is critical to understand and interpret VUS in patient context for utmost utility. Utilizing large language models and automated systems for VUS reclassification is on the rise. This will certainly impact several aspects including patient management, treatment surveillance, prophylactic opportunities, and preventing disease inheritance. Here, we describe the utility of our machine model system for effective interpretation and reclassification of VUS in patient context to reduce uncertainty and achieve high clinical utility. Methods CGP was performed on patients using next-generation sequencing (NGS) with the OncoIndx® panel. Variant reclassification was performed through our in-house automated precision classification and interpretation system. Results Using our computational molecular oncology workflow, a precision classification and interpretation system has been developed for the confirmation of variants of uncertain significance (VUS). The system when tested on several VUS intronic based on in silico evidence (PP5, PM2, BP4) and low evolutionary conservation scores (-0.423), ended up being reclassified as likely pathogenic. The system also investigates for the presence of conditions like Lynch-syndrome and its associated genomic findings including high microsatellite instability, loss of MSH2 protein on IHC. Functional consequences of the computational predictions such as weakening of the native acceptor site, and functional RNA studies are also utilized to confirm functionality. Finally, the patient and family history act as critical parameters. Using these combined evidence, the precision classification system generates a final verdict of reclassification. Conclusion When complementary lines of evidence pertaining to functional loss of a variant (PS3), its identification in affected individuals (PS4_supporting), segregation-consistent family history (PP1), rarity in population databases (PM2), and supportive in silico predictions (PP3/BP4)-are integrated, a coherent and biologically consistent explanation can be generated for disease causation. Our automated machine system in-built with robust and essential criteria for variant classification, thus can be optimally utilized to reclassifiy VUS and improve clinical outcomes.