An Ensemble Approach to Predict the Pathogenicity of Synonymous Variants

Quantifying negative selection on synonymous variants

Widespread adoption of DNA sequencing has resulted in large numbers of genetic variants, whose contribution to disease is not easily determined. Although many types of variation are known to disrupt cellular processes in predictable ways, for some categories of variants, the effects may not be directly detectable. A particular example is synonymous variants, that is, those single-nucleotide variants that create a codon substitution, such that the produced amino acid sequence is unaffected. Contrary to the original theory suggesting that synonymous variants are benign, there is a growing volume of research showing that, despite their "silent" mechanism of action, some synonymous variation may be deleterious. Here, we studied the extent of the negative selective pressure acting on different classes of synonymous variants by analyzing the relative enrichment of synonymous singleton variants in the human exomes provided by gnomAD. Using a modification of the mutability-adjusted proportion of singletons (MAPS) metric as a measure of purifying selection, we found that some classes of synonymous variants are subject to stronger negative selection than others. For instance, variants that reduce codon optimality undergo stronger selection than optimality-increasing variants. Besides, selection affects synonymous variants implicated in splice-site-loss or splice-site-gain events. To understand what drives this negative selection, we tested a number of predictors in the aim to explain the variability in the selection scores. Our findings provide insights into the effects of synonymous variants at the population level, highlighting the specifics of the role that these variants play in health and disease.

Nonsynonymous Synonymous Variants Demand for a Paradigm Shift in Genetics

Synonymous (also known as silent) variations are by definition not considered to change the coded protein. Still many variations in this category affect either protein abundance or properties. As this situation is confusing, we have recently introduced systematics for synonymous variations and those that may on the surface look like synonymous, but these may affect the coded protein in various ways. A new category, unsense variation, was introduced to describe variants that do not introduce a stop codon into the variation site, but which lead to different types of changes in the coded protein. Many of these variations lead to mRNA degradation and missing protein. Here, consequences of the systematics are discussed from the perspectives of variation annotation and interpretation, evolutionary calculations, nonsynonymous-to-synonymous substitution rates, phylogenetics and other evolutionary inferences that are based on the principle of (nearly) neutral synonymous variations. It may be necessary to reassess published results. Further, databases for synonymous variations and prediction methods for such variations should consider unsense variations. Thus, there is a need to evaluate and reflect principles of numerous aspects in genetics, ranging from variation naming and classification to evolutionary calculations.

AmazonForest: In Silico Metaprediction of Pathogenic Variants

Simple Summary

ClinVar is a valuable platform that stores a large set of relevant genetic associations with complex phenotypes. However, the functional impact of a partial set of such associations remains misinterpreted, due to the presence of variants with uncertain significance or with conflicting pathogenicity interpretations. To fill this gap, we present AmazonForest: a metaprediction model based on Random Forest for pathogenicity prediction. AmazonForest was used to reclassify a set of ∼101,000 variants that were predicted as having high pathogenic probability. AmazonForest is available as a web tool with a simple web interface, and also as an R object for pathogenicity predictions.

Abstract

ClinVar is a web platform that stores ∼789,000 genetic associations with complex diseases. A partial set of these cataloged genetic associations has challenged clinicians and geneticists, often leading to conflicting interpretations or uncertain clinical impact significance. In this study, we addressed the (re)classification of genetic variants by AmazonForest, which is a random-forest-based pathogenicity metaprediction model that works by combining functional impact data from eight prediction tools. We evaluated the performance of representation learning algorithms such as autoencoders to propose a better strategy. All metaprediction models were trained with ClinVar data, and genetic variants were annotated with eight functional impact predictors cataloged with SnpEff/SnpSift. AmazonForest implements the best random forest model with a one hot data-encoding strategy, which shows an Area Under ROC Curve of ≥0.93. AmazonForest was employed for pathogenicity prediction of a set of ∼101,000 genetic variants of uncertain significance or conflict of interpretation. Our findings revealed ∼24,000 variants with high pathogenic probability (RFprob≥0.9). In addition, we show results for Alzheimer’s Disease as a demonstration of its application in clinical interpretation of genetic variants in complex diseases. Lastly, AmazonForest is available as a web tool and R object that can be loaded to perform pathogenicity predictions.