Boosting In-Silicon Directed Evolution with Fine-Tuned Protein Language Model and Tree Search

2025-11-13

AI Summary

Current silicon-based directed evolution algorithms fail to effectively leverage evolutionary priors encoded in protein language models (PLMs). To address this, we propose PLM-MCTS: a framework that first fine-tunes a pre-trained PLM on homologous sequences via masked language modeling, then integrates test-time inference with Monte Carlo Tree Search (MCTS) to enable interpretable, guided exploration of evolutionary trajectories. This work represents the first systematic incorporation of large-scale PLMs into the directed evolution pipeline, enabling efficient compression of sequence space within a limited number of mutational steps. On multiple protein family benchmarks, PLM-MCTS significantly outperforms state-of-the-art methods in functional sequence discovery, demonstrating that evolutionary patterns implicitly learned by PLMs provide effective, data-driven guidance for computational protein evolution.

Technology Category

• Machine Learning: Evolutionary Learning
• Search and Optimization: Evolutionary Computation
• Planning, Routing, and Scheduling: Planning with Language Models

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Cost models of using LLMs in production systems
• User Modeling, Personalization and Recommendation: Large Language Models (LLM) for user modeling and recommendation

Abstract

Protein evolution through amino acid sequence mutations is a cornerstone of life sciences. While current in-silicon directed evolution algorithms focus on designing search strategies, they overlook how to utilize the transformative protein language models, which encode rich evolutionary patterns, to guide search. To bridge this gap, we propose AlphaDE, a novel framework to evolve protein sequences by harnessing the innovative paradigms of large language models. First, AlphaDE fine-tunes pretrained protein language models using masked language modeling on homologous protein sequences to activate the evolutionary plausibility for the interested protein class. Second, AlphaDE introduces test-time inference based on Monte Carlo tree search, which effectively evolves proteins with evolutionary guidance from the fine-tuned protein language model. Extensive benchmark experiments show that AlphaDE remarkably outperforms previous state-of-the-art methods even with few-shot fine-tuning. An interesting case study further shows that AlphaDE supports condensing the protein sequence space through computational evolution.

Problem

• Utilizing protein language models to guide evolutionary search strategies
• Evolving protein sequences with evolutionary plausibility and guidance
• Condensing protein sequence space through computational directed evolution

Innovation

• Fine-tunes protein language models using masked language modeling
• Introduces Monte Carlo tree search for evolutionary guidance
• Condenses protein sequence space through computational evolution