ProDCARL: Reinforcement Learning-Aligned Diffusion Models for De Novo Antimicrobial Peptide Design

2026-01-29

AI Summary

This work addresses the challenge of explicitly optimizing both high antimicrobial activity and low toxicity in antimicrobial peptide (AMP) design—a limitation of conventional generative models—by introducing, for the first time, a reinforcement learning alignment mechanism into a diffusion model. Building upon the EvoDiff OA-DM 38M architecture, the proposed framework employs AMP activity and toxicity predictors to guide sequence generation, integrating top-k policy gradients, entropy regularization, and early stopping to enhance performance while preserving sequence diversity. Experimental results demonstrate a significant improvement in generated peptides, with the average AMP score increasing from 0.081 to 0.178. Notably, 6.3% of candidates exhibit high quality (pAMP > 0.7 and pTox < 0.3), and sequence diversity remains high at 0.929, effectively balancing potency, safety, and novelty.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation
• Search and Optimization: Sampling/Simulation-based Search

Application Category

• User Modeling, Personalization and Recommendation: Attacks and countermeasures in recommendation systems
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking

Abstract

Antimicrobial resistance threatens healthcare sustainability and motivates low-cost computational discovery of antimicrobial peptides (AMPs). De novo peptide generation must optimize antimicrobial activity and safety through low predicted toxicity, but likelihood-trained generators do not enforce these goals explicitly. We introduce ProDCARL, a reinforcement-learning alignment framework that couples a diffusion-based protein generator (EvoDiff OA-DM 38M) with sequence property predictors for AMP activity and peptide toxicity. We fine-tune the diffusion prior on AMP sequences to obtain a domain-aware generator. Top-k policy-gradient updates use classifier-derived rewards plus entropy regularization and early stopping to preserve diversity and reduce reward hacking. In silico experiments show ProDCARL increases the mean predicted AMP score from 0.081 after fine-tuning to 0.178. The joint high-quality hit rate reaches 6.3\% with pAMP $>$0.7 and pTox $<$0.3. ProDCARL maintains high diversity, with $1-$mean pairwise identity equal to 0.929. Qualitative analyses with AlphaFold3 and ProtBERT embeddings suggest candidates show plausible AMP-like structural and semantic characteristics. ProDCARL serves as a candidate generator that narrows experimental search space, and experimental validation remains future work.

Problem

• antimicrobial peptides
• de novo design
• toxicity prediction
• antimicrobial activity
• computational discovery

Innovation

• diffusion models
• reinforcement learning alignment
• de novo peptide design
• antimicrobial peptides
• reward hacking mitigation