Unraveling the Potential of Diffusion Models in Small Molecule Generation
arXiv:2507.08005 · doi:10.1016/j.drudis.2025.104413
Abstract
Generative AI presents chemists with novel ideas for drug design and facilitates the exploration of vast chemical spaces. Diffusion models (DMs), an emerging tool, have recently attracted great attention in drug R\&D. This paper comprehensively reviews the latest advancements and applications of DMs in molecular generation. It begins by introducing the theoretical principles of DMs. Subsequently, it categorizes various DM-based molecular generation methods according to their mathematical and chemical applications. The review further examines the performance of these models on benchmark datasets, with a particular focus on comparing the generation performance of existing 3D methods. Finally, it concludes by emphasizing current challenges and suggesting future research directions to fully exploit the potential of DMs in drug discovery.
References in corpus (39)
- Denoising Diffusion Probabilistic Models
- Training language models to follow instructions with human feedback
- Diffusion Models Beat GANs on Image Synthesis
- Generative Modeling by Estimating Gradients of the Data Distribution
- GuacaMol: Benchmarking Models for De Novo Molecular Design
- Junction Tree Variational Autoencoder for Molecular Graph Generation
- Classifier-Free Diffusion Guidance
- DiffDock: Diffusion Steps, Twists, and Turns for Molecular Docking
- SE(3)-Transformers: 3D Roto-Translation Equivariant Attention Networks
- Diffusion-LM Improves Controllable Text Generation
- GeoDiff: a Geometric Diffusion Model for Molecular Conformation Generation
- Equivariant Diffusion for Molecule Generation in 3D
- Flow Matching for Generative Modeling
- Torsional Diffusion for Molecular Conformer Generation
- Analog Bits: Generating Discrete Data using Diffusion Models with Self-Conditioning
- 3D Equivariant Diffusion for Target-Aware Molecule Generation and Affinity Prediction
- DiGress: Discrete Denoising diffusion for graph generation
- Generating 3D Molecules for Target Protein Binding
- SILVR: Guided Diffusion for Molecule Generation
- Geometric Latent Diffusion Models for 3D Molecule Generation
- Consistency Models
- DecompDiff: Diffusion Models with Decomposed Priors for Structure-Based Drug Design
- MUDiff: Unified Diffusion for Complete Molecule Generation
- Geometric-Facilitated Denoising Diffusion Model for 3D Molecule Generation
- MolDiff: Addressing the Atom-Bond Inconsistency Problem in 3D Molecule Diffusion Generation
- Functional-Group-Based Diffusion for Pocket-Specific Molecule Generation and Elaboration
- Shape-conditioned 3D Molecule Generation via Equivariant Diffusion Models
- DecompOpt: Controllable and Decomposed Diffusion Models for Structure-based Molecular Optimization
- Swallowing the Bitter Pill: Simplified Scalable Conformer Generation
- Navigating the Design Space of Equivariant Diffusion-Based Generative Models for De Novo 3D Molecule Generation
- Benchmarking structure-based three-dimensional molecular generative models using GenBench3D: ligand conformation quality matters
- Autoregressive fragment-based diffusion for pocket-aware ligand design
- Re-Dock: Towards Flexible and Realistic Molecular Docking with Diffusion Bridge
- 3D molecule generation by denoising voxel grids
- Coarse-to-Fine: a Hierarchical Diffusion Model for Molecule Generation in 3D
- Aligning Target-Aware Molecule Diffusion Models with Exact Energy Optimization
- AUTODIFF: Autoregressive Diffusion Modeling for Structure-based Drug Design
- STRIDE: Structure-guided Generation for Inverse Design of Molecules
- E(3)-equivariant models cannot learn chirality: Field-based molecular generation