Torsional Diffusion for Molecular Conformer Generation

2022/06/01 by Bowen Jing, Jing, Bowen, Gabriele Corso +7 · 46 citations
Biochemistry, Genetics and Molecular Biology · Mathematics · #Bioinformatics and Genomic Networks #Biomolecules (q-bio.BM) #Chemical Physics (physics.chem-ph) #FOS: Biological sciences #FOS: Computer and information sciences #FOS: Physical sciences #Machine Learning (cs.LG) #Mathematical Biology Tumor Growth #Protein Structure and Dynamics

paper · pdf · doi:10.48550/arxiv.2206.01729

openalex publication_date 2022/06/01 · openalex created_date 2025/10/10 · openalex updated_date 2026/07/28

Abstract

Molecular conformer generation is a fundamental task in computational chemistry. Several machine learning approaches have been developed, but none have outperformed state-of-the-art cheminformatics methods. We propose torsional diffusion, a novel diffusion framework that operates on the space of torsion angles via a diffusion process on the hypertorus and an extrinsic-to-intrinsic score model. On a standard benchmark of drug-like molecules, torsional diffusion generates superior conformer ensembles compared to machine learning and cheminformatics methods in terms of both RMSD and chemical properties, and is orders of magnitude faster than previous diffusion-based models. Moreover, our model provides exact likelihoods, which we employ to build the first generalizable Boltzmann generator. Code is available at https://github.com/gcorso/torsional-diffusion.

Cited by

• Energy-Guided Flow Matching Enables Few-Step Conformer Generation and Ground-State Identification
• In-Context Semi-Supervised Learning
• Unified Control for Inference-Time Guidance of Denoising Diffusion Models
• MolSculpt: Sculpting 3D Molecular Geometries from Chemical Syntax
• FlexiFlow: decomposable flow matching for generation of flexible molecular ensemble
• Chemistry-Enhanced Diffusion-Based Framework for Small-to-Large Molecular Conformation Generation
• VEDA: 3D Molecular Generation via Variance-Exploding Diffusion with Annealing
• Recurrent Equivariant Constraint Modulation: Learning Per-Layer Symmetry Relaxation from Data
• Protein Language Model Embeddings Improve Generalization of Implicit Transfer Operators
• HollowFlow: Efficient Sample Likelihood Evaluation using Hollow Message Passing
• Learning Boltzmann Generators via Constrained Mass Transport
• FLOWR.root: A flow matching based foundation model for joint multi-purpose structure-aware 3D ligand generation and affinity prediction
• A Conformation-Centric Generative Foundation Model for Linear Polymer Modeling and Design
• Discrete State Diffusion Models: A Sample Complexity Perspective
• Transferable Generative Models Bridge Femtosecond to Nanosecond Time-Step Molecular Dynamics
• Flow-Matching Based Refiner for Molecular Conformer Generation
• MolSpectLLM: A Molecular Foundation Model Bridging Spectroscopy, Molecule Elucidation, and 3D Structure Generation
• Guiding Diffusion Models with Reinforcement Learning for Stable Molecule Generation
• Feynman-Kac-Flow: Inference Steering of Conditional Flow Matching to an Energy-Tilted Posterior
• Diffusion Language Models Know the Answer Before Decoding
• Energy-Based Flow Matching for Generating 3D Molecular Structure
• SynCoGen: Synthesizable 3D Molecule Generation via Joint Reaction and Coordinate Modeling
• Torsional-GFN: a conditional conformation generator for small molecules
• Efficient Molecular Conformer Generation with SO(3)-Averaged Flow Matching and Reflow
• Geometric Generative Modeling with Noise-Conditioned Graph Networks
• Kinetic Langevin Diffusion for Crystalline Materials Generation
• BoltzNCE: Learning Likelihoods for Boltzmann Generation with Stochastic Interpolants and Noise Contrastive Estimation
• Unraveling the Potential of Diffusion Models in Small Molecule Generation
• A standard transformer and attention with linear biases for molecular conformer generation
• Sampling 3D Molecular Conformers with Diffusion Transformers
• PDESpectralRefiner: Achieving More Accurate Long Rollouts with Spectral Adjustment
• Rethinking Losses for Diffusion Bridge Samplers
• Wasserstein Convergence of Score-based Generative Models under Semiconvexity and Discontinuous Gradients
• Efficient Diffusion Models for Symmetric Manifolds
• Beyond Equilibrium: Non-Equilibrium Foundations Should Underpin Generative Processes in Complex Dynamical Systems
• Improved Sample Complexity For Diffusion Model Training Without Empirical Risk Minimizer Access
• Applications of Modular Co-Design for De Novo 3D Molecule Generation
• Flexible MOF Generation with Torsion-Aware Flow Matching
• Minimum-Excess-Work Guidance
• Path Gradients after Flow Matching
• InvDesFlow-AL: active learning-based workflow for inverse design of functional materials
• ItDPDM: Information-Theoretic Discrete Poisson Diffusion Model
• Flow Sampling: Learning to Sample from Unnormalized Densities via Denoising Conditional Processes
• polyGen: A Learning Framework for Atomic-level Polymer Structure Generation
• Efficient Training of Boltzmann Generators Using Off-Policy Log-Dispersion Regularization
• PolyConf: Unlocking Polymer Conformation Generation through Hierarchical Generative Models

Related

• Torsional-GFN: a conditional conformation generator for small molecules
• Torsion-Space Diffusion for Protein Backbone Generation with Geometric Refinement
• Mitigating Exposure Bias in Score-Based Generation of Molecular Conformations
• EC-Conf: An Ultra-fast Diffusion Model for Molecular Conformation Generation with Equivariant Consistency
• DiffPack: A Torsional Diffusion Model for Autoregressive Protein Side-Chain Packing