Atomic Diffusion Models for Small Molecule Structure Elucidation from NMR Spectra

2025-12-02

AI Summary

This work addresses the heavy reliance on expert knowledge in small-molecule structural elucidation by proposing ChefNMR—the first end-to-end, NMR-spectrum-driven method for fully automated 3D structure generation. ChefNMR introduces a novel conditional generative framework that integrates an atomic diffusion model with a non-equivariant Transformer, enabling direct prediction of 3D conformations from only 1D ¹H/¹³C NMR spectra and molecular formula. To enable robust training, the authors curate a high-fidelity simulated NMR dataset comprising over 110,000 natural products. On a challenging natural-product test set, ChefNMR achieves >65% Top-1 structural accuracy—substantially outperforming existing approaches. The code and dataset are publicly released, establishing a scalable, reproducible paradigm for automated molecular structure determination.

Technology Category

• Machine Learning: Structured Learning
• Natural Language Processing: Code Generation / Program Synthesis from Natural Language
• Knowledge Representation and Reasoning: Automated Reasoning and Theorem Proving

Application Category

• Graph Algorithms and Modeling for the Web: Foundation models and LLMs for Web-related graphs
• Web Mining and Content Analysis: Large pretrained models with web data
• Semantics and Knowledge: Methods, algorithms and applications for the development of semantic models, knowledge graphs and other forms of structured data models with machine-interpretable semantics

Abstract

Nuclear Magnetic Resonance (NMR) spectroscopy is a cornerstone technique for determining the structures of small molecules and is especially critical in the discovery of novel natural products and clinical therapeutics. Yet, interpreting NMR spectra remains a time-consuming, manual process requiring extensive domain expertise. We introduce ChefNMR (CHemical Elucidation From NMR), an end-to-end framework that directly predicts an unknown molecule's structure solely from its 1D NMR spectra and chemical formula. We frame structure elucidation as conditional generation from an atomic diffusion model built on a non-equivariant transformer architecture. To model the complex chemical groups found in natural products, we generated a dataset of simulated 1D NMR spectra for over 111,000 natural products. ChefNMR predicts the structures of challenging natural product compounds with an unsurpassed accuracy of over 65%. This work takes a significant step toward solving the grand challenge of automating small-molecule structure elucidation and highlights the potential of deep learning in accelerating molecular discovery. Code is available at https://github.com/ml-struct-bio/chefnmr.

Problem

• Automates small molecule structure prediction from NMR spectra
• Reduces manual interpretation time and expertise dependency
• Enhances accuracy in natural product compound identification

Innovation

• Atomic diffusion model for NMR-based structure prediction
• Non-equivariant transformer architecture for conditional generation
• Simulated NMR dataset of 111,000 natural products