UniMoMo: Unified Generative Modeling of 3D Molecules for De Novo Binder Design

2025-03-25

AI Summary

Existing generative models are confined to single molecular domains (e.g., small molecules or antibodies only), limiting their applicability to multi-target therapeutics and preventing cross-domain knowledge transfer. This work introduces the first unified 3D generative framework for binders across small molecules, peptides, and antibodies. Our method employs: (1) a domain-agnostic graph-patch representation using amino acids or chemical fragments as atomic units; (2) an E(3)-equivariant geometric latent diffusion process coupled with an iterative all-atom autoencoder; and (3) joint multi-domain training and zero-shot cross-domain generalization. Evaluated on three comprehensive benchmarks, our approach consistently outperforms state-of-the-art single-domain models, achieving significant improvements in generation fidelity, structural diversity, and target-binding compatibility. The framework establishes a general-purpose generative paradigm for structure-based drug discovery.

Technology Category

• Machine Learning: Transfer, Domain Adaptation, Multi-Task Learning
• Computer Vision: Diffusion Models for Vision
• Natural Language Processing: Generation

Application Category

• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Economics, Online Markets and Human Computation: Economic ramifications for generative AI infrastructure and applications

Abstract

The design of target-specific molecules such as small molecules, peptides, and antibodies is vital for biological research and drug discovery. Existing generative methods are restricted to single-domain molecules, failing to address versatile therapeutic needs or utilize cross-domain transferability to enhance model performance. In this paper, we introduce Unified generative Modeling of 3D Molecules (UniMoMo), the first framework capable of designing binders of multiple molecular domains using a single model. In particular, UniMoMo unifies the representations of different molecules as graphs of blocks, where each block corresponds to either a standard amino acid or a molecular fragment. Based on these unified representations, UniMoMo utilizes a geometric latent diffusion model for 3D molecular generation, featuring an iterative full-atom autoencoder to compress blocks into latent space points, followed by an E(3)-equivariant diffusion process. Extensive benchmarks across peptides, antibodies, and small molecules demonstrate the superiority of our unified framework over existing domain-specific models, highlighting the benefits of multi-domain training.

Problem

• Design target-specific molecules for drug discovery
• Overcome single-domain limitations in generative methods
• Unify molecular representations for cross-domain transferability

Innovation

• Unified graph representation for multi-domain molecules
• Geometric latent diffusion model for 3D generation
• Iterative full-atom autoencoder for block compression