Quantitative Missense Variant Effect Prediction Using Large-Scale Mutagenesis Data

Abstract

Large datasets describing the quantitative effects of mutations on protein function are becoming increasingly available. Here, we leverage these datasets to develop Envision, which predicts the magnitude of a missense variant's molecular effect. Envision combines 21,026 variant effect measurements from nine large-scale experimental mutagenesis datasets, a hitherto untapped training resource, with a supervised, stochastic gradient boosting learning algorithm. Envision outperforms other missense variant effect predictors both on large-scale mutagenesis data and on an independent test dataset comprising 2,312 TP53 variants whose effects were measured using a low-throughput approach. This dataset was never used for hyperparameter tuning or model training and thus serves as an independent validation set. Envision prediction accuracy is also more consistent across amino acids than other predictors. Finally, we demonstrate that Envision's performance improves as more large-scale mutagenesis data are incorporated. We precompute Envision predictions for every possible single amino acid variant in human, mouse, frog, zebrafish, fruit fly, worm, and yeast proteomes (https://envision.gs.washington.edu/).

A resource of variant effect predictions of single nucleotide variants in model organisms

“…As additional data become available, mutfunc will be updated to improve coverage and future work could expand the set of mechanisms studied such as drug or small‐molecule binding sites, RNA‐binding interfaces, among others. The effects of variants on molecular and cellular phenotypes are increasingly being probed directly by large‐scale mutagenesis experiments (Fowler & Fields, ; Weile et al , ), which will likely result in improved variant effect prediction algorithms (Gray et al , ). The curation of such experimentally determined effects and the improved algorithms can be integrated in future iterations of mutfunc.…”

Assessment of methods for predicting the effects of PTEN and TPMT protein variants

“…In addition, Envision, a machine learning model trained on large‐scale mutagenesis data sets (similar to this study) to predict the impact of missense variants on protein function, was also included as a newer baseline method (Gray, Hause, Luebeck, Shendure, & Fowler, ). No score transformation was necessary for Envision.…”

“…While these were not developed for the specific task of predicting effects of variants on protein stability, the VSP assay itself was originally developed as a general-purpose technique for the high-throughput identification of deleterious and benign variants and may thus be somewhat comparable with the effects these methods were developed to predict. For the baseline methods, score transformation was carried out (heuristically) to match the expected score range as follows In addition, Envision, a machine learning model trained on large-scale mutagenesis data sets (similar to this study) to predict the impact of missense variants on protein function, was also included as a newer baseline method (Gray, Hause, Luebeck, Shendure, & Fowler, 2018). No score transformation was necessary for Envision.…”

fuNTRp: Identifying protein positions for variation driven functional tuning

“…Thus, it is infeasible to experimentally assess, for example, the effects of all non-synonymous Single Nucleotide Polymorphisms (nsSNPs) of a given individual, much less a population. However, the large-scale mutational fitness landscapes resulting from DMS analyses are an exciting resource for the development of new accurate variant effect prediction approaches (Gray, et al, 2018).…”