Rigidity-Aware Geometric Pretraining for Protein Design and Conformational Ensembles

2026-03-02

AI Summary

Current protein generative models struggle to simultaneously capture global rigid geometry and dynamic conformational ensembles, limiting their performance in design and conformation prediction. This work proposes RigidSSL, a novel framework that introduces, for the first time, a rigidity-aware bidirectional flow matching objective to jointly optimize translational and rotational dynamics, thereby unifying the modeling of protein geometry and dynamics. The approach employs a two-stage self-supervised pretraining strategy: first learning geometric priors from the AlphaFold database, followed by representation refinement using molecular dynamics trajectories. This significantly enhances performance—boosting protein design success rates by up to 43%, markedly improving novelty and diversity in unconditional generation, increasing zero-shot motif scaffolding success by 5.8%, and yielding GPCR conformational ensembles that better align with biophysical reality.

Technology Category

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

Application Category

• Search and Retrieval-Augmented AI: Retrieval-Augmented Generation (RAG) and multi-modal RAG
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Uses of LLMs and GenAI for marketplace design, bidding, and strategic interactions

Abstract

Generative models have recently advanced $\textit{de novo}$ protein design by learning the statistical regularities of natural structures. However, current approaches face three key limitations: (1) Existing methods cannot jointly learn protein geometry and design tasks, where pretraining can be a solution; (2) Current pretraining methods mostly rely on local, non-rigid atomic representations for property prediction downstream tasks, limiting global geometric understanding for protein generation tasks; and (3) Existing approaches have yet to effectively model the rich dynamic and conformational information of protein structures. To overcome these issues, we introduce $\textbf{RigidSSL}$ ($\textit{Rigidity-Aware Self-Supervised Learning}$), a geometric pretraining framework that front-loads geometry learning prior to generative finetuning. Phase I (RigidSSL-Perturb) learns geometric priors from 432K structures from the AlphaFold Protein Structure Database with simulated perturbations. Phase II (RigidSSL-MD) refines these representations on 1.3K molecular dynamics trajectories to capture physically realistic transitions. Underpinning both phases is a bi-directional, rigidity-aware flow matching objective that jointly optimizes translational and rotational dynamics to maximize mutual information between conformations. Empirically, RigidSSL variants improve designability by up to 43\% while enhancing novelty and diversity in unconditional generation. Furthermore, RigidSSL-Perturb improves the success rate by 5.8\% in zero-shot motif scaffolding and RigidSSL-MD captures more biophysically realistic conformational ensembles in G protein-coupled receptor modeling. The code is available at: https://github.com/ZhanghanNi/RigidSSL.git.

Problem

• protein design
• conformational ensembles
• geometric pretraining
• rigidity-aware learning
• generative modeling

Innovation

• Rigidity-Aware Learning
• Geometric Pretraining
• Protein Conformational Ensembles
• Self-Supervised Learning
• Flow Matching