Low-N Protein Activity Optimization with FolDE

2025-10-28

AI Summary

Protein directed evolution under low-data regimes suffers from high experimental costs, while existing active learning–driven approaches (e.g., ALDE) exhibit poor generalization due to greedy, homogeneous batch selection. To address these limitations, we propose FolDE—a framework that synergistically integrates protein language model (PLM) priors with scarce experimental data. FolDE introduces a naturalness-aware pretraining strategy and a “perpetual-lie” batch selector, jointly promoting mutational diversity and prediction accuracy. Comprehensive evaluation across 20 protein targets demonstrates that FolDE significantly outperforms state-of-the-art methods: it improves the discovery rate of the top 10% most active variants by 23% and increases the probability of identifying top 1% variants by 55%. These gains underscore FolDE’s enhanced sample efficiency and robustness in low-data optimization scenarios.

Technology Category

• Machine Learning: Active Learning
• Search and Optimization: Sampling/Simulation-based Search
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• User Modeling, Personalization and Recommendation: Federated recommendation systems and personalization

Abstract

Proteins are traditionally optimized through the costly construction and measurement of many mutants. Active Learning-assisted Directed Evolution (ALDE) alleviates that cost by predicting the best improvements and iteratively testing mutants to inform predictions. However, existing ALDE methods face a critical limitation: selecting the highest-predicted mutants in each round yields homogeneous training data insufficient for accurate prediction models in subsequent rounds. Here we present FolDE, an ALDE method designed to maximize end-of-campaign success. In simulations across 20 protein targets, FolDE discovers 23% more top 10% mutants than the best baseline ALDE method (p=0.005) and is 55% more likely to find top 1% mutants. FolDE achieves this primarily through naturalness-based warm-starting, which augments limited activity measurements with protein language model outputs to improve activity prediction. We also introduce a constant-liar batch selector, which improves batch diversity; this is important in multi-mutation campaigns but had limited effect in our benchmarks. The complete workflow is freely available as open-source software, making efficient protein optimization accessible to any laboratory.

Problem

• Optimizing protein activity with limited experimental data
• Overcoming training data homogeneity in active learning evolution
• Improving mutant discovery efficiency using computational methods

Innovation

• FolDE uses naturalness-based warm-starting for prediction
• It employs constant-liar batch selector for diversity
• Combines protein language model outputs with measurements