Learning protein fitness landscapes with deep mutational scanning data from multiple sources

One of the key points of machine learning-assisted directed evolution (MLDE) is the accurate learning of the fitness landscape, a conceptual mapping from sequence variants to the desired function. Here, we describe a multi-protein training scheme to leverage the existing deep mutational scanning data from different proteins to help reason about the fitness landscape of a new protein. Proof-of-concept trials are designed to validate this training scheme in three aspects: random and positional extrapolation for single variant effects, zero-shot fitness predictions for new proteins, and extrapolation for higher-order variant effects from single variant effects. Moreover, our study identified previously overlooked strong baselines, and their unexpectedly good performance bring our attention to the pitfalls of MLDE. Overall, these results may improve our understanding of the association between different protein fitness profiles, and shed light on developing better machine learning-assisted approaches to the directed evolution of proteins.

Two zipped files were deposited here. "directed_evolution_input_datasets.tar.gz" contains the processed DMS datasets and "zero-shot_parameter.tar.gz" contains our pretrained model parameters for zero-shot prediction.