Bioinformatics Seminar
Time: 11AM
Venue: Zoom Webinar
22 February 2022
Integrating deep mutational scanning and low-throughput mutagenesis data into models of variant effect prediction
James FuWEHI Bioinformatics
The availability of whole-genome sequencing has led to the discovery of numerous variants of uncertain significance (VUS). Experimental methods like deep mutational scanning (DMS) can measure the functional impact of large number of mutations on a target protein and help to evaluate the impact of these VUS; however, DMS data are not available for the majority of proteins. Multiple studies have therefore sought to model DMS data computationally to predict the functional consequences of variants on unassayed proteins. In this study, we explore whether further incorporating alanine scanning (AS) data improves the performance of these DMS models. We collected 135 alanine scanning results and 54 DMS datasets targeting 20 distinct proteins and find improved model performance with an extended model when the DMS and AS data are highly compatible. We also show that the scale of improvement is closely related to the DMS and AS score correlation. Further investigations explore extensions to other kinds of mutagenesis data and variant effect prediction models.