Multi-state Protein Design with DynamicMPNN

2025-07-29

AI Summary

Current multi-state protein design methods rely on posterior aggregation of single-state predictions, resulting in low experimental success rates and poor generation of sequences compatible with multiple conformations. To address this, we propose DynamicMPNN—the first model enabling joint conformational modeling and end-to-end sequence design, breaking the conventional single-sequence–single-structure paradigm. DynamicMPNN employs a conformation-set joint training strategy and incorporates AlphaFold-based initial prediction evaluation. It is trained on 46,033 conformation pairs spanning 75% of CATH superfamilies. On multi-state benchmarks, it achieves a 13% reduction in normalized RMSD over ProteinMPNN, significantly improving structural accuracy and experimental realizability. This advances precise modeling and design of dynamic biological processes—such as enzyme catalysis and membrane transport—establishing a new paradigm for conformationally adaptive protein engineering.

Technology Category

• Machine Learning: Large Multimodal Models (LMMs)
• Natural Language Processing: Language Grounding & Multi-modal NLP
• Computer Vision: Multi-modal Vision

Application Category

• Graph Algorithms and Modeling for the Web: Efficient manipulation of static and dynamic Web-related graphs
• User Modeling, Personalization and Recommendation: On-Device user modeling, personalization, and recommendation
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Structural biology has long been dominated by the one sequence, one structure, one function paradigm, yet many critical biological processes - from enzyme catalysis to membrane transport - depend on proteins that adopt multiple conformational states. Existing multi-state design approaches rely on post-hoc aggregation of single-state predictions, achieving poor experimental success rates compared to single-state design. We introduce DynamicMPNN, an inverse folding model explicitly trained to generate sequences compatible with multiple conformations through joint learning across conformational ensembles. Trained on 46,033 conformational pairs covering 75% of CATH superfamilies and evaluated using AlphaFold initial guess, DynamicMPNN outperforms ProteinMPNN by up to 13% on structure-normalized RMSD across our challenging multi-state protein benchmark.

Problem

• Designing proteins with multiple conformational states
• Improving multi-state sequence prediction accuracy
• Enhancing compatibility across conformational ensembles

Innovation

• Joint learning across conformational ensembles
• Inverse folding model for multi-conformations
• Outperforms ProteinMPNN by 13%