Membrane proteins are the primary means by which cells interact and communicate with their environment. Despite their central importance, our understanding of membrane proteins lags significantly behind soluble proteins, in large part due to experimental challenges in studying them. In particular, we do not understand the systematic rules that govern how changes to DNA sequence impact membrane protein expression. Understanding the determinants of membrane protein expression (also known as biogenesis) is crucial, as many of the disease-causing mutations in membrane proteins alter protein expression. Recent work from a number of labs, including our own, has enabled the high-throughput characterization of all possible point mutations across different membrane proteins using deep mutational scanning (DMS). These studies have revealed important determinants of protein expression and function within each individual protein, but the extent to which these findings generalize across membrane proteins remains unclear. To address this gap, we assembled a comprehensive catalogue of 20 distinct DMS data sets across 19 proteins, including previously published as well as unpublished data sets, all of which measured the impact of point mutants on membrane protein expression. We then processed all of these data sets through a single feature extraction pipeline we created that calculated sequence-based, structure-based, and biophysical features. Using this harmonized feature representation, we evaluated which features were able to predict membrane protein expression. We then compared the types of features that came up as predictive in GPCRs, transporters, and ion channels to evaluate which features were specific to distinct classes of membrane proteins. Finally, we trained machine learning models of increasingly complexity on this data to evaluate our ability to predict the impact of mutations on expression. All of the underlying code, data, and trained models will be open-sourced to serve as a community resource.
Greenwald et al. (Sun,) studied this question.