Abstracts Category : Other

Predicting the Effects of Protein Variants using Structural Modeling, Large-Scale Data Integration, and Machine Learning

by Evan H Baugh

High-throughput sequencing technologies and new computational techniques for analyzing population genetics data are rapidly improving our understanding of disease susceptibility in humans and adaptation in a wide variety of organisms. These studies often discover nonsynonymous variation with large effects as even a single amino acid change can disrupt the folding, catalytic activity, and physical interactions of proteins. Current estimates predict that every human genome contains 10,000-11,000 nonsynonymous variations and, while we cannot currently characterize all this diversity experimentally, many variants that alter protein function can be identified computationally from destabilization of structural models or amino acid conservation. Methods for annotating variant effects in genome-wide association studies and exome sequencing studies use conservation and other sequence-based features to identify damaging variants but cannot predict the effect these variants have on protein function. Recent studies of de novo variants have demonstrated the power of these methods but also the need for additional information, such as physical models from the Protein Data Bank, to identify causal variants in disease association studies. I present VIPUR, a computational framework that integrates sequence analysis and structural modeling using the Rosetta protein modeling suite to identify and interpret deleterious protein variants. To train VIPUR, I collected 9,477 protein variants with known effects on protein function from multiple organisms and curated structural models for each variant from crystal structures and homology models. VIPUR can be applied to variants in any organisms proteome with improved generalized accuracy (AUROC .83) and interpretability (AUPR .87) compared to other methods. I show that VIPURs predictions of deleteriousness match the biological phenotypes for pathogenicity in ClinVar despite being trained on a different label. I use VIPUR to interpret mutations associated with inflammation and diabetes, demonstrating the structural diversity of disrupted functional sites and improved interpretation functional effects. Generalizable tools for interpreting genetic variants are especially needed with individualized exome sequencing, where clear indications of confident predictions are necessary to identify causal variation. I demonstrate VIPURs ability to select candidate variants associated with human diseases by predicting the effects of <i>de novo</i> variants associated with Autism Spectrum Disorders (ASD) in the Simons Simplex Collection. Compared to existing methods, VIPUR deleterious predictions have the greatest enrichment for mutations found in children with ASD. VIPURs predictions of deleterious effects are easily combined with other protein functional data to produce a small set of candidate genes and variants with specific mechanistic predictions. Although designed to aid in the discovery of causal variants, VIPUR can also simulate mutations