AI-guided discovery for low-resource peptide engineering using evolutionary scale modeling
- 1Division of Molecular Bioscience, Department of Life Sciences, Chalmers University of Technology
Abstract
Data scarcity is a major bottleneck in artificial intelligence (AI)-guided peptide engineering, where experimental training sets often contain only tens to a few hundred sequences, and it is difficult to judge in advance whether a protein language model (PLM) will effectively guide iterative discovery before committing to wet-lab work. Here we present SCARSE, a machine learning framework that couples ESM2 protein language model embeddings with Gaussian process regression. Benchmarked across 24 diverse peptide and small-protein datasets spanning substitution and indel variants, antimicrobial peptides, and hemolytic peptides, SCARSE significantly outperformed hand-engineered descriptor baselines on substitution and indel tasks in low-data regimes (20-500 training samples) and performed comparably on short non-mutant peptides. In active learning simulations it consistently prioritized high-performing candidates, discovering up to 7-fold more top-tier peptides than random sampling. Cross-validation (CV) R2 correlated strongly with final active learning performance (Pearson r = 0.80-0.84), and evaluating on as few as 50 labeled peptides reduced CV R2 variance enough for it to serve as a practical, data-efficient criterion for assessing model suitability before iterative screening. SCARSE thus provides an early diagnostic indicator for active learning success and is freely available as an open-source Python package (pip) and through an interactive Hugging Face Spaces web interface.
Related
- iAMP-deep: an alignment-free AI framework for insect antimicrobial peptide discovery
- False discovery rate control enables trustworthy deep-learning-based de novo peptide sequencing
- A Plasmodium falciparum Pangenome Resource to Drive Structural Variant Discovery and to assist Malaria Control
- An eco-evolutionary consumer-resource theory of host-microbe symbioses
- ALSmiR: a curated evidence resource for ALS-associated microRNAs for biomarker discovery
BibTeX
@misc{andrekson2026aiguided,
title = {{AI-guided discovery for low-resource peptide engineering using evolutionary scale modeling}},
author = {L. Andrekson and R. Rydbergh and R. Mercado and M. Wenzel},
howpublished = {bioRxiv},
year = {2026},
month = jun,
doi = {10.64898/2026.06.25.734678},
url = {https://www.biorxiv.org/content/10.64898/2026.06.25.734678v2},
note = {CC BY},
}
RIS
TY - UNPB
TI - AI-guided discovery for low-resource peptide engineering using evolutionary scale modeling
AU - L. Andrekson
AU - R. Rydbergh
AU - R. Mercado
AU - M. Wenzel
PY - 2026
DA - 2026/06/25
PB - bioRxiv
DO - 10.64898/2026.06.25.734678
UR - https://www.biorxiv.org/content/10.64898/2026.06.25.734678v2
AB - Data scarcity is a major bottleneck in artificial intelligence (AI)-guided peptide engineering, where experimental training sets often contain only tens to a few hundred sequences, and it is difficult to judge in advance whether a protein language model (PLM) will effectively guide iterative discovery before committing to wet-lab work. Here we present SCARSE, a machine learning framework that couples ESM2 protein language model embeddings with Gaussian process regression. Benchmarked across 24 diverse peptide and small-protein datasets spanning substitution and indel variants, antimicrobial peptides, and hemolytic peptides, SCARSE significantly outperformed hand-engineered descriptor baselines on substitution and indel tasks in low-data regimes (20-500 training samples) and performed comparably on short non-mutant peptides. In active learning simulations it consistently prioritized high-performing candidates, discovering up to 7-fold more top-tier peptides than random sampling. Cross-validation (CV) R2 correlated strongly with final active learning performance (Pearson r = 0.80-0.84), and evaluating on as few as 50 labeled peptides reduced CV R2 variance enough for it to serve as a practical, data-efficient criterion for assessing model suitability before iterative screening. SCARSE thus provides an early diagnostic indicator for active learning success and is freely available as an open-source Python package (pip) and through an interactive Hugging Face Spaces web interface.
ER -