A General Harness for Protein Foundation Model Fitness Prediction
Abstract
Accurate fitness prediction is central to protein engineering and understanding sequence-function relationships. With advances in deep learning, protein foundation models (PFMs) have become widely used for this task. Recent analyses, however, show that these models share preferences reflecting their training corpora, while unreliable inputs can further distort fitness predictions. Family-specific evolutionary evidence and structural context can help address these limitations by providing complementary constraints on model scores, motivating VenusREM-Harness (VRH), a general, model-agnostic, training-free Retrieval-Enhanced Mutation harness. It fuses frozen model scores with multiple sequence alignment (MSA) evidence according to model uncertainty, then applies gated background correction and score shrinkage based on structural confidence and solvent exposure. Across 1,211 assays and 3.1 million measured variants from ProteinGym, VenusMutHub, and the newly curated viral benchmark VenusViroHub, all 71 configurations improve Spearman correlation on all 3 benchmarks by 0.073 on average, with broad gains across 5 metrics. Extended analyses relate retrieval gains to model-MSA preference differences, assess domain-level gains and immune-escape cases, and quantify computational speedups. Built with VRH, VenusREM2 is the first to rank highest in all function, taxon, MSA-depth, and mutation-depth categories, with a ProteinGym Average Spearman of 0.556, 0.038 above the prior best.
Related
- From single-sequence structure prediction to protein fitness landscape through a composable, epistasis-aware mutation atlas
- Generalizing Protein Fitness Prediction Across Distribution Shifts via Structural and Co-evolutionary Priors
- Limitations of general protein language models and public TCRpMHC data in training specificity prediction models
- Filling the Missing Puzzle: Integrating Interaction and Evolutionary Information to Enhance Protein Fitness Prediction in Inverse Folding
- Decision benchmarking reveals when protein foundation models add experimental value
BibTeX
@misc{tan2026general,
title = {{A General Harness for Protein Foundation Model Fitness Prediction}},
author = {Yang Tan and Qijia Tian and Gangyu Sun and Bozitao Zhong and Mingchen Li and Yuanxi Yu and Nanqing Dong and Liang Hong},
howpublished = {arXiv},
year = {2026},
month = sep,
doi = {10.48550/arxiv.2609.34654},
eprint = {2609.34654},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2609.34654},
}
RIS
TY - UNPB
TI - A General Harness for Protein Foundation Model Fitness Prediction
AU - Yang Tan
AU - Qijia Tian
AU - Gangyu Sun
AU - Bozitao Zhong
AU - Mingchen Li
AU - Yuanxi Yu
AU - Nanqing Dong
AU - Liang Hong
PY - 2026
DA - 2026/09/28
PB - arXiv
DO - 10.48550/arxiv.2609.34654
UR - https://arxiv.org/abs/2609.34654
AB - Accurate fitness prediction is central to protein engineering and understanding sequence-function relationships. With advances in deep learning, protein foundation models (PFMs) have become widely used for this task. Recent analyses, however, show that these models share preferences reflecting their training corpora, while unreliable inputs can further distort fitness predictions. Family-specific evolutionary evidence and structural context can help address these limitations by providing complementary constraints on model scores, motivating VenusREM-Harness (VRH), a general, model-agnostic, training-free Retrieval-Enhanced Mutation harness. It fuses frozen model scores with multiple sequence alignment (MSA) evidence according to model uncertainty, then applies gated background correction and score shrinkage based on structural confidence and solvent exposure. Across 1,211 assays and 3.1 million measured variants from ProteinGym, VenusMutHub, and the newly curated viral benchmark VenusViroHub, all 71 configurations improve Spearman correlation on all 3 benchmarks by 0.073 on average, with broad gains across 5 metrics. Extended analyses relate retrieval gains to model-MSA preference differences, assess domain-level gains and immune-escape cases, and quantify computational speedups. Built with VRH, VenusREM2 is the first to rank highest in all function, taxon, MSA-depth, and mutation-depth categories, with a ProteinGym Average Spearman of 0.556, 0.038 above the prior best.
ER -