Fitness inference from deep mutational scanning data

Deep mutational scanning (DMS) is a popular method used to test the functional effects of mutations at very large scales. DMS experiments generate large amounts of data that can be difficult to analyze, with large variation between experimental replicates. Here, we combined methods from statistical physics and population genetics to interpret the functional effects of mutations from DMS data. We applied our approach to DMS data from both human proteins and viruses. Through extensive tests, we find that our method infers substantially more consistent functional effects of mutations than current state-of-the-art methods, while also being computationally efficient. Our pipeline can be widely applied to DMS data including multiple time points, replicates, and conditions.