MAXWELL: Calibrating the probabilistic outputs of protein language models to the mutation-induced stability change landscape
- 1School of Physics and Astronomy & Institute of Natural Sciences, Shanghai Jiao Tong University, Shanghai 200240, China; Zhangjiang Institute for Advanced Study,
Abstract
Designing mutations that enhance protein stability is a central goal in protein engineering. However, experimentally screening large numbers of candidate mutations is costly and time-consuming, creating a strong need for computational methods that can identify potentially stabilizing mutations. Among these approaches, protein language models are particularly promising because they learn context-dependent amino acid preferences from large-scale sequence and structure datasets. Nevertheless, most existing stability prediction methods use these models primarily as feature extractors and do not fully exploit the amino acid probability distributions they encode. Here, we introduce MAXWELL (Matrix-wise Landscape Learning), a novel post-training method that calibrates the probabilistic outputs learned by protein language models during pretraining to generate mutational landscapes that quantify the effects of individual amino acid substitutions on protein stability. When applied to ProteinMPNN, MAXWELL yields a state-of-the-art predictor of the effects of protein mutations on stability, outperforming ThermoMPNN and other representative methods on a curated benchmark of experimentally measured stability changes. We next applied MAXWELL to the design of ten single-point mutations in the DhaA dehalogenase, seven of which (70%) increased thermal stability. Among them, G171W showed the largest improvement, with a measured {Delta}T of 4.91 . These experimental results establish MAXWELL as a novel post-training strategy for protein language models and a practical framework for designing stabilizing mutations.
Related
- Decoding antibody-antigen recognition: the BL8 language and its application in calibrating deep learning predictions
- From single-sequence structure prediction to protein fitness landscape through a composable, epistasis-aware mutation atlas
- Towards Sparse Causal Features for Zero-shot Mutation Effect Prediction in a Protein Language Model
- Probing Physical Intuition in Large Language Models through Partner-Dependent Mutation Prediction
- Unified modeling of mutation-induced protein stability responses with OmegaTherm
BibTeX
@misc{li2026maxwell,
title = {{MAXWELL: Calibrating the probabilistic outputs of protein language models to the mutation-induced stability change landscape}},
author = {M. Li and X. Cheng and M. Hong and F. Jiang and L. Hong and Y. Yu},
howpublished = {bioRxiv},
year = {2026},
month = jul,
doi = {10.64898/2026.07.29.741631},
url = {https://www.biorxiv.org/content/10.64898/2026.07.29.741631v2},
note = {CC BY-NC-ND},
}
RIS
TY - UNPB
TI - MAXWELL: Calibrating the probabilistic outputs of protein language models to the mutation-induced stability change landscape
AU - M. Li
AU - X. Cheng
AU - M. Hong
AU - F. Jiang
AU - L. Hong
AU - Y. Yu
PY - 2026
DA - 2026/07/29
PB - bioRxiv
DO - 10.64898/2026.07.29.741631
UR - https://www.biorxiv.org/content/10.64898/2026.07.29.741631v2
AB - Designing mutations that enhance protein stability is a central goal in protein engineering. However, experimentally screening large numbers of candidate mutations is costly and time-consuming, creating a strong need for computational methods that can identify potentially stabilizing mutations. Among these approaches, protein language models are particularly promising because they learn context-dependent amino acid preferences from large-scale sequence and structure datasets. Nevertheless, most existing stability prediction methods use these models primarily as feature extractors and do not fully exploit the amino acid probability distributions they encode. Here, we introduce MAXWELL (Matrix-wise Landscape Learning), a novel post-training method that calibrates the probabilistic outputs learned by protein language models during pretraining to generate mutational landscapes that quantify the effects of individual amino acid substitutions on protein stability. When applied to ProteinMPNN, MAXWELL yields a state-of-the-art predictor of the effects of protein mutations on stability, outperforming ThermoMPNN and other representative methods on a curated benchmark of experimentally measured stability changes. We next applied MAXWELL to the design of ten single-point mutations in the DhaA dehalogenase, seven of which (70%) increased thermal stability. Among them, G171W showed the largest improvement, with a measured {Delta}T of 4.91 . These experimental results establish MAXWELL as a novel post-training strategy for protein language models and a practical framework for designing stabilizing mutations.
ER -