DeMaSk: a deep mutational scanning substitution matrix and its use for variant impact prediction

• DOI: 10.1093/bioinformatics/btaa1030
• Publication Date: 2020-11-30T20:30:03Z
• Authors: Daniel Munro; Mona Singh
• Keywords: Leverage (statistics); Substitution (logic); Amino acid substitution; Protein sequencing; Sequence (biology)

Accurately predicting the quantitative impact of a substitution on protein's molecular function would be great aid in understanding effects observed genetic variants across populations. While this remains challenging task, new approaches can leverage data from increasing numbers comprehensive deep mutational scanning (DMS) studies that systematically mutate proteins and measure fitness.We introduce DeMaSk, an intuitive interpretable method based only upon DMS datasets sequence homologs predicts missense mutations within any protein. DeMaSk first infers directional amino acid matrix then fits linear model combines these scores with measures per-position evolutionary conservation variant frequency homologs. Despite its simplicity, has state-of-the-art performance substitutions, easily rapidly applied to protein sequence.https://demask.princeton.edu generates fitness predictions visualizations for user-submitted sequence.Supplementary are available at Bioinformatics online.