Functional Assessment of Genetic Variants with Outcomes Adapted to Clinical Decision-Making

Understanding the medical effect of an ever-growing number of human variants detected is a long term challenge in genetic counseling. Functional assays, based on in vitro or in vivo evaluations of the variant effects, provide essential information, but they require robust statistical validation, as well as adapted outputs, to be implemented in the clinical decision-making process. Here, we assessed 25 pathogenic and 15 neutral missense variants of the BRCA1 breast/ovarian cancer susceptibility gene in four BRCA1 functional assays. Next, we developed a novel approach that refines the variant ranking in these functional assays. Lastly, we developed a computational system that provides a probabilistic classification of variants, adapted to clinical interpretation. Using this system, the best functional assay exhibits a variant classification accuracy estimated at 93%. Additional theoretical simulations highlight the benefit of this ready-to-use system in the classification of variants after functional assessment, which should facilitate the consideration of functional evidences in the decision-making process after genetic testing. Finally, we demonstrate the versatility of the system with the classification of siRNAs tested for human cell growth inhibition in high throughput screening.Author Summary: Human genetics has entered a new age with the advent of next generation sequencing. However, this great advance also comes with new concerns. Currently, the extensive use of multi-gene panels, whole exome and whole genome sequencing, is generating an ever-growing number of new DNA sequence variations detected in the disease-predisposing genes of human patients. The pathogenic or neutral status of these variants needs to be known before planning any medical act or follow-up. We show here that the status of the variants identified in the BRCA1 breast/ovarian cancer susceptibility gene can be estimated thanks to experimental systems using yeast cells and a novel computational model. Importantly, this model provides a probabilistic classification of variants, opening the possibility to integrate results from functional assays into clinical decision-making. Moreover, our computational model is directly compatible with all kinds of experimental system without any requirement for skills in statistics thanks to ready-to-use online tools. We believe that this work is a step forward in the clinical interpretation of human genetic variants.