Inferring the molecular and phenotypic impact of amino acid variants with MutPred2.

Pejaver V, Urresti J, Lugo-Martinez J, Pagel KA, Lin GN, Nam HJ, Mort M, Cooper DN, Sebat J, Iakoucheva LM, Mooney SD, Radivojac P. · 2020

Read Inferring the molecular and phenotypic impact of amino acid variants with MutPred2. by Pejaver V, Urresti J, Lugo-Martinez J, Pagel KA, Lin GN, Nam HJ, Mort M, Cooper DN, Sebat J, Iakoucheva LM, Mooney SD, Radivojac P. (2020) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc7680112.txt.

Introduction

Pejaver and co-authors introduce a tool that predicts how single protein changes affect health. The tool, MutPred2, scores amino acid swaps in proteins. It aims to spot which rare variants cause disease. It also guesses the exact protein feature each variant disrupts.

The tool ranks harmful variants better than past methods. Its scores come with clear, testable ideas about mechanism. The authors used it on disorders passed down in families and on new variants linked to brain development problems. Lab tests then backed several of its calls in patients. They argue such mechanism-first tools can speed the hunt for treatment targets.

Catalog record

CREATORPejaver V, Urresti J, Lugo-Martinez J, Pagel KA, Lin GN, Nam HJ, Mort M, Cooper DN, Sebat J, Iakoucheva LM, Mooney SD, Radivojac P.
YEAR2020
COLLECTIONPAPERS
FORMATTXT
FILE SIZE78 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC7680112

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Inferring the molecular and phenotypic impact of amino acid variants with MutPred2 By Vikas Pejaver, Jorge Urresti, Jose Lugo-Martinez, Kymberleigh A Pagel, Guan Ning Lin, Hyun-Jun Nam, Matthew Mort, David N Cooper, Jonathan Sebat, Lilia M Iakoucheva, Sean D Mooney, Predrag Radivojac Published in: Nature Communications DOI: 10.1038/s41467-020-19669-x ABSTRACT Abstract Identifying pathogenic variants and underlying functional alterations is challenging. To this end, we introduce MutPred2, a tool that improves the prioritization of pathogenic amino acid substitutions over existing methods, generates molecular mechanisms potentially causative of disease, and returns interpretable pathogenicity score distributions on individual genomes. Whilst its prioritization performance is state-of-the-art, a distinguishing feature of MutPred2 is the probabilistic modeling of variant impact on specific aspects of protein structure and function that can serve to guide experimental studies of phenotype-altering variants. We demonstrate the utility of MutPred2 in the identification of the structural and functional mutational signatures relevant to Mendelian disorders and the prioritization of de novo mutations associated with complex neurodevelopmental disorders. We then experimentally validate the functional impact of several variants identified in patients with such disorders. We argue that mechanism-driven studies of human inherited disease have the potential to significantly accelerate the discovery of clinically actionable variants. Subject terms: Computational biology and bioinformatics, Machine learning, Protein analysis, Protein function predictions, Protein structure predictions INTRODUCTION The discovery of pathogenic variants, i.e., variants capable of causing disease, generally relies on a combination of family- and population-based sequencing efforts1. To assist genetic studies, particularly in characterizing rare variants and dissecting complex disease, machine learning methods have recently been developed to identify the signatures of pathogenicity and to predict the impact of variants of unknown significance2,3. Although pathogenicity prediction methods have matured considerably over the past decade and are now routinely integrated into genomic pipelines, they continue to exhibit major shortcomings. Firstly, they remain inadequate to the task in exome-scale applications owing to a less than optimal balance of false-positive and true-positive detection rates4,5. Secondly, they do not generate actionable hypotheses regarding the molecular consequences of these variants6. The functional impact of variants may lead to a wide range of molecular changes, even within a single protein, including disrupted stability and structure, disrupted macromolecular binding, ablation of posttranslational modification (PTM) sites, among others (Fig. 1a). However, existing approaches generally provide little or no information about the potential mechanisms affected by mutations, or else simply map predicted pathogenic substitutions onto protein feature annotations (which are generally sparse) in public databases. These methods do not therefore explicitly model the type of change in local structure and function, and are fundamentally limited by the incomplete, incorrect, and inevitably biased annotations in major databases7–9. [Figure: MutPred2 and the molecular consequences of amino acid substitutions.a The human tumor suppressor p53 as an illustration of the numerous possible effects of amino acid substitutions on protein structure and function. Protein Data Bank IDs for the structures shown are 1TUP, 1YCS, 2J1W, and 2YBG. b The ontology constructed in this study to organize the possible structural and functional effects of amino acid substitutions. It is confined to the 53 properties included in MutPred2. c The MutPred2 workflow. For a given amino acid sequence and substitution, MutPred2 first extracts six categories of features. Changes in structure and function due to the substitution are also modeled by running the original and mutated sequences through different sequence-based protein property predictors. Two scores are obtained for each property and these are combined to generate two additional scores quantifying the loss and gain of the property in question. All four scores are included as features. Next, all categories of features are presented to an ensemble of 30 neural networks trained to distinguish between pathogenic and benign variants. MutPred2 returns two outputs, the general score and the property score. The general score is obtained from the neural network ensemble and indicates the pathogenicity of the given variant. It ranges between 0 and 1, with a higher score indicating a greater propensity to be pathogenic. The property score is assigned to each of the 53 properties for the given variant and also ranges between 0 and 1. The latter score is the posterior probability of loss or gain (whichever is greater) of the given property due to the substitution. The higher the property score, the more likely that the molecular mechanism of the disease involves the alteration of the property.] To address these challenges, we have extended our existing and widely used machine learning approach, MutPred10, by developing an improved and statistically rigorous approach, MutPred2. This new algorithm quantifies the pathogenicity of amino acid substitutions, and describes how they affect the phenotype by modeling a broad repertoire of structural and functional alterations from amino acid sequence. MutPred2 compares favorably with the existing tools recommended in the American College of Medical Genetics and Genomics and the Association for Molecular Pathology (ACMG/AMP) Standards and Guidelines11 on stringent independent test sets. More importantly, by applying this methodology, we estimate the fraction of pathogenic missense variants in a personal genome, and identify molecular signatures associated with a data set of Mendelian disease variants and a data set of de novo mutations found in individuals diagnosed with neurodevelopmental disorders. Finally, we prioritize several high-scoring variants from this data set and experimentally validate their functional roles. Our results suggest new molecular targets and mechanisms impacted by multiple mutations across neurodevelopmental disorders. More broadly, this study demonstrates the power of the proposed mechanism-driven approach to studying human phenotypes. RESULTS Overview of MutPred2 MutPred2 is a machine learning-based method, and software package that integrates genetic and molecular data to reason probabilistically about the pathogenicity of amino acid substitutions. This is achieved by providing (1) a general pathogenicity prediction, and (2) a ranked list of specific molecular alterations potentially affecting the phenotype. MutPred2 is a sequence-based model that utilizes methodology predicated upon recent machine learning advances in training from positive-unlabeled data, and which incorporates estimation of prior and posterior probabilities12,13. These estimates not only facilitate the interpretation of pathogenicity and molecular alteration scores, but also provide a framework to rigorously rank the underlying mechanisms13. Currently, MutPred2 models a broad range of structural and functional properties, including secondary structure, signal peptide and transmembrane topology, catalytic activity, macromolecular binding, PTMs, metal binding and allostery. Challenges in variant interpretation tool development To develop models for the mathematically sound inference of molecular mechanisms of disease, several statistical and computational challenges must be addressed. First, it is necessary to integrate disparate molecular and genetic data to develop models that have similar yet meaningful score interpretations6. Second, prediction software tools generally vary not only in terms of their feature representation and prediction algorithms, but also in their implementations, dependencies, and system requirements, which collectively hinder the development of a robust framework that seamlessly incorporates multiple models. Third, structural and functional properties occur with unequal prior probabilities, requiring sophisticated modeling to rank the properties affected by a substitution. Finally, although these property predictors are typically developed independently of one another, they are interrelated; i.e., a single substitution may affect more than one property. This places the burden of interpretation upon the user and can be overwhelming, when multiple properties are considered simultaneously. To address the first two challenges, we developed sequence-based predictors

Read the complete text