Guiding Generative Protein Language Models with Reinforcement Learning

2024/12/17 by Filippo Stocco, Maria Artigues-Lleixa, Stocco, Filippo +10 · 1 voice · 3 citations
Computer Science · Biochemistry, Genetics and Molecular Biology · #Topic Modeling #Machine Learning in Bioinformatics #Natural Language Processing Techniques

paper · pdf · doi:10.48550/arxiv.2412.12979

Abstract

Protein language models (pLMs) have demonstrated success at generating functional proteins across vast sequence spaces but lack the ability to design high-fitness variants on demand. Here, we iteratively guide pLMs toward user-defined objectives by applying reinforcement learning (RL). We demonstrate that RL can steer pLMs toward various protein properties, such as topologies or binding affinities, in a few iterations through long evolutionary trajectories. We apply our framework to the design of epidermal growth factor receptor (EGFR) binders, achieving a 26-fold increase in binding affinity in two iterations.

Cited by

• Generative AI for Enzyme Design and Biocatalysis
• Guiding Generative Models for Protein Design: Prompting, Steering and Aligning
• AnnoDPO: Protein Functional Annotation Learning with Direct Preference Optimization

Discussions

• Two more🙃: 1) arxiv.org/abs/2412.12979 2) arxiv.org/abs/2505.15093 [bsky, 15 points, 0 comments]

Related

• From Supervision to Exploration: What Does Protein Language Model Learn During Reinforcement Learning?
• Reinforcement Learning for Sequence Design Leveraging Protein Language Models
• Robust Optimization in Protein Fitness Landscapes Using Reinforcement Learning in Latent Space
• Guiding Generative Models for Protein Design: Prompting, Steering and Aligning
• ProteinZero: Self-Improving Protein Generation via Online Reinforcement Learning