Protein generation with embedding learning for motif diversification

2025-10-21

AI Summary

Protein design faces a fundamental trade-off between structural diversity and preservation of functional motifs: existing methods—such as partial diffusion in RFdiffusion—either converge to similar structures under small perturbations or violate geometric constraints critical for function under large perturbations. To address this, we propose PGEL, the first framework that learns and controllably perturbs functional motif representations within the frozen high-dimensional embedding space of a diffusion denoiser, thereby bypassing destructive geometric distortions inherent in coordinate-space perturbations. PGEL jointly encodes sequence and structural features, enabling functional-constrained structural diversification directly at the embedding layer. Experiments on monomer design, protein–protein interfaces, and oncology-relevant transcription factor complexes demonstrate that PGEL significantly improves structural diversity, designability, and self-consistency—outperforming partial-diffusion baselines across all metrics.

Technology Category

• Machine Learning: Structured Learning
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Learning to Search

Application Category

• Graph Algorithms and Modeling for the Web: Representation, reconstruction, and subgraph or motif discovery in Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Economics, Online Markets and Human Computation: Incentives in network design for Web infrastructures and ecosystems

Abstract

A fundamental challenge in protein design is the trade-off between generating structural diversity while preserving motif biological function. Current state-of-the-art methods, such as partial diffusion in RFdiffusion, often fail to resolve this trade-off: small perturbations yield motifs nearly identical to the native structure, whereas larger perturbations violate the geometric constraints necessary for biological function. We introduce Protein Generation with Embedding Learning (PGEL), a general framework that learns high-dimensional embeddings encoding sequence and structural features of a target motif in the representation space of a diffusion model's frozen denoiser, and then enhances motif diversity by introducing controlled perturbations in the embedding space. PGEL is thus able to loosen geometric constraints while satisfying typical design metrics, leading to more diverse yet viable structures. We demonstrate PGEL on three representative cases: a monomer, a protein-protein interface, and a cancer-related transcription factor complex. In all cases, PGEL achieves greater structural diversity, better designability, and improved self-consistency, as compared to partial diffusion. Our results establish PGEL as a general strategy for embedding-driven protein generation allowing for systematic, viable diversification of functional motifs.

Problem

• Balancing structural diversity with motif function preservation
• Overcoming limitations in geometric constraint satisfaction
• Generating viable protein structures with enhanced diversity

Innovation

• Learns embeddings encoding protein sequence and structure features
• Introduces controlled perturbations in embedding space for diversity
• Loosens geometric constraints while maintaining biological function viability