Composable Score-based Graph Diffusion Model for Multi-Conditional Molecular Generation

2025-09-11

AI Summary

Existing controllable molecular graph generation methods suffer from limited performance under multi-condition joint control, primarily due to reliance on joint conditional modeling or continuous relaxations, which compromise structural fidelity. This paper introduces the first score-matching diffusion model explicitly designed for discrete molecular graph spaces, enabling flexible, subset-wise conditional guidance over arbitrary combinations of molecular attributes. Key contributions include: (1) the first extension of score matching to discrete molecular graph spaces; (2) a Composable Guidance (CoG) mechanism for fine-grained, decoupled conditional control; and (3) a Probability Calibration (PC) strategy ensuring training–inference consistency. Evaluated on four standard benchmarks, our method achieves an average 15.3% improvement in controllability while maintaining >98.7% molecular validity and superior distributional fidelity.

Technology Category

• Machine Learning: Graph-based Machine Learning
• Computer Vision: Diffusion Models for Vision
• Search and Optimization: Mixed Discrete/Continuous Search

Application Category

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

Abstract

Controllable molecular graph generation is essential for material and drug discovery, where generated molecules must satisfy diverse property constraints. While recent advances in graph diffusion models have improved generation quality, their effectiveness in multi-conditional settings remains limited due to reliance on joint conditioning or continuous relaxations that compromise fidelity. To address these limitations, we propose Composable Score-based Graph Diffusion model (CSGD), the first model that extends score matching to discrete graphs via concrete scores, enabling flexible and principled manipulation of conditional guidance. Building on this foundation, we introduce two score-based techniques: Composable Guidance (CoG), which allows fine-grained control over arbitrary subsets of conditions during sampling, and Probability Calibration (PC), which adjusts estimated transition probabilities to mitigate train-test mismatches. Empirical results on four molecular datasets show that CSGD achieves state-of-the-art performance, with a 15.3% average improvement in controllability over prior methods, while maintaining high validity and distributional fidelity. Our findings highlight the practical advantages of score-based modeling for discrete graph generation and its capacity for flexible, multi-property molecular design.

Problem

• Multi-conditional molecular graph generation with property constraints
• Addressing limited effectiveness in multi-conditional graph diffusion models
• Enabling flexible control over arbitrary condition subsets during sampling

Innovation

• Concrete scores enable discrete graph diffusion
• Composable Guidance allows fine-grained condition control
• Probability Calibration mitigates train-test mismatches