Learning Protein Fitness Landscapes with Multimodal Stability Priors

Shannon SQ Zhang, Yunan Luo

Keywords: protein engineering, protein function prediction, protein language models, multi-modal foundation models, protein stability, protein fitness

TL;DR: PsiFit improves low-N protein fitness prediction by integrating structural stability priors into protein language model adaptation.

Abstract: Predicting how mutations change protein fitness is central to protein engineering and variant interpretation, yet most experimentally measured fitness landscapes contain only limited labeled variants. We present PsiFit, a stability-informed framework that adapts protein language models for low-$N$ fitness prediction by injecting mutation-induced stability changes predicted by a multimodal sequence-structure foundation model. By integrating biophysical stability priors into contrastive fine-tuning, PsiFit aims to improve data efficiency, reduce overfitting, and provide a general strategy for learning protein fitness landscapes from sparse assays.

Email Sharing: We authorize the sharing of all author emails with Program Chairs.

Data Release: We authorize the release of our submission and author names to the public in the event of acceptance.

Submission Number: 107