MODA: A Unified 3D Diffusion Framework for Multi-Task Target-Aware Molecular Generation

2025-07-09

AI Summary

Existing 3D molecular diffusion generators suffer from SMILES-only input, a two-stage pretraining-finetuning paradigm, and task-specific architectures, resulting in poor stereochemical fidelity, weak task alignment, and limited zero-shot transferability. Method: We propose the first single-stage unified 3D diffusion framework capable of joint generation across multiple tasks—including fragment growth, linker design, scaffold hopping, and side-chain modification—and introduce a Bayesian masking scheduler that enables end-to-end learning of shared chemical and geometric priors, eliminating reliance on force-field optimization. Contribution/Results: Our model consistently outperforms six baseline methods in substructure matching, property prediction, intermolecular interaction modeling, and geometric accuracy. In zero-shot de novo design, it achieves stable negative Vina scores and high structural optimization success rates. The framework significantly enhances generalizability and task alignment while unifying diverse generative objectives within a single architecture.

Technology Category

• Computer Vision: Diffusion Models for Vision
• Machine Learning: Deep Neural Architectures and Foundation Models
• Natural Language Processing: Generation

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Search and Retrieval-Augmented AI: Web learning to rank, online learning, and counterfactual learning for ranking
• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs

Abstract

Three-dimensional molecular generators based on diffusion models can now reach near-crystallographic accuracy, yet they remain fragmented across tasks. SMILES-only inputs, two-stage pretrain-finetune pipelines, and one-task-one-model practices hinder stereochemical fidelity, task alignment, and zero-shot transfer. We introduce MODA, a diffusion framework that unifies fragment growing, linker design, scaffold hopping, and side-chain decoration with a Bayesian mask scheduler. During training, a contiguous spatial fragment is masked and then denoised in one pass, enabling the model to learn shared geometric and chemical priors across tasks. Multi-task training yields a universal backbone that surpasses six diffusion baselines and three training paradigms on substructure, chemical property, interaction, and geometry. Model-C reduces ligand-protein clashes and substructure divergences while maintaining Lipinski compliance, whereas Model-B preserves similarity but trails in novelty and binding affinity. Zero-shot de novo design and lead-optimisation tests confirm stable negative Vina scores and high improvement rates without force-field refinement. These results demonstrate that a single-stage multi-task diffusion routine can replace two-stage workflows for structure-based molecular design.

Problem

• Fragmented molecular generators lack stereochemical fidelity and task alignment
• Current methods hinder zero-shot transfer and require task-specific models
• Need unified framework for multi-task molecular generation with shared priors

Innovation

• Unified diffusion framework for multi-task molecular generation
• Bayesian mask scheduler enables shared geometric priors
• Single-stage training replaces two-stage workflows