arXiv 2602.087522026-02cond-mat.mtrl-sci

NMRPeak：一种基于协同跨模态学习的分子结构解析智能系统

NMRPeak: a ready-to-use intelligent system for molecular structure elucidation enabled by synergistic cross-modal learning

Institute of Artificial Intelligence, Xiamen University（厦门大学人工智能研究院）；State Key Laboratory of Physical Chemistry of Solid Surfaces, iChEM, College of Chemistry and Chemical Engineering, Xiamen University（厦门大学化学化工学院）；DP Technology（星环科技）；Laboratory of AI for Electrochemistry (AI4EC), Tan Kah Kee Innovation Laboratory (IKKEM)（陈嘉庚创新实验室电化学人工智能实验室）；College of Chemistry and Molecular Engineering, Peking University（北京大学化学与分子工程学院）；Pen-Tung Sah Institute of Micro-Nano Science and Technology, Discipline of Intelligent Instrument and Equipment, iChEM, Xiamen University（厦门大学微纳科学技术研究院）；AI for Science Institute（人工智能科学研究机构）

Fanjie Xu, Jinyuan Hu, Jingxiang Zou, Junjie Wang, Boying Huang, Zhifeng Gao, Xiaohong Ji, Weinan E, Zhong-Qun Tian, Fujie Tang, Jun Cheng

备注：: 35 pages, 7 figures

阅读 PDF

arXiv 页面 ↗HTML 原文 ↗

AI总结：

NMRPeak通过实验驱动设计整合谱预测、检索和生成任务，实现分子结构解析的突破性性能，提升实验场景下的应用效用。

摘要：

一维核磁共振（NMR）光谱在有机合成、药物发现、天然产物表征和代谢组学中至关重要，但其解释仍严重依赖专家知识且难以扩展。尽管机器学习已应用于NMR光谱预测、库检索和结构生成，但这些任务孤立地使用模拟数据和不兼容的光谱表示，限制了其在真实实验场景中的实用性。本文介绍了NMRPeak，一种统一的跨模态学习系统，通过实验驱动设计整合这三个任务。我们整理了约180万个实验和模拟光谱，构建了最大的NMR结构解析基准，并系统量化了这些领域的分布偏移。我们引入了化学意识的自适应分词器，动态平衡离散化粒度以保留光谱语义同时控制词汇量，并引入无分配的峰感知相似度度量，使预测和实验光谱之间能够直接比较。通过统一的分子到光谱范式和预测、检索和生成模块的协同耦合，NMRPeak在实验基准上实现了突破性性能：克服了长期存在的模拟到实验光谱预测差距，同时在分子检索中达到超过95%的top-1准确率，在立体化学意识的de novo结构生成中达到约75%的top-1准确率。这些能力为有机合成、药物发现和化学生物学中的自动化、高通量分子结构解析奠定了基础。

英文摘要：

One-dimensional nuclear magnetic resonance (NMR) spectroscopy is essential for molecular structure elucidation in organic synthesis, drug discovery, natural product characterization, and metabolomics, yet its interpretation remains heavily dependent on expert knowledge and difficult to scale. Although machine learning has been applied to NMR spectrum prediction, library retrieval, and structure generation, these tasks have evolved in isolation using simulated data and incompatible spectral representations, limiting their utility under real experimental scenarios. Here we present NMRPeak, a unified cross-modal learning system that integrates these three tasks through experimentally grounded design. We curate approximately 1.8 million experimental and simulated spectra to construct the largest benchmark for NMR-based structure elucidation and systematically quantify the distribution shift between these domains. We introduce a chemically-aware adaptive tokenizer that dynamically balances discretization granularity to preserve spectral semantics while controlling vocabulary size, and an assignment-free peak-aware similarity metric that enables direct comparison between predicted and experimental spectra. Through a unified molecule-to-spectrum paradigm and synergistic coupling of prediction, retrieval, and generation modules, NMRPeak achieves transformative performance on experimental benchmarks: it overcomes the longstanding simulation-to-experiment gap in spectrum prediction while delivering over 95% top-1 accuracy in molecular retrieval and approximately 75% top-1 accuracy in stereochemistry-aware de novo structure generation. These capabilities establish a foundation for automated, high-throughput molecular structure elucidation in organic synthesis, drug discovery, and chemical biology.

机构、总结、中文摘要均由AI分析而来，可能存在错误，请以论文正文为准。