TCM-MS2Link: A Unified AI-Ready Dataset Integrating TCM Herb–Compound Knowledge and MS/MS Spectral Data

This study presents TCM-MS2Link, a standardized mass spectrometry-based association dataset for traditional Chinese medicine (TCM), serving as an important resource for natural product research in TCM. The dataset adopts a dual-layer “knowledge–data” architecture: the first layer, TCM-MolLink, comprises curated herb–compound association data, constructed through the integration of multiple heterogeneous databases and rigorous consistency filtering to establish high-confidence relationships between TCM herbs and their chemical constituents; the second layer, MS2-MLReady, is a benchmark dataset for mass spectrometry-based machine learning which, after systematic data cleaning, standardized preprocessing, and well-designed data partitioning, can directly support the training and evaluation of artificial intelligence models. By addressing key limitations in existing public resources, including data fragmentation, inconsistent annotations, and insufficient computational usability, TCM-MS2Link effectively overcomes major bottlenecks in the systematic analysis of TCM components and data-driven research. This study significantly enhances the reliability of herb–compound associations and the modeling readiness of mass spectrometry data, providing a high-quality, standardized, and reusable data foundation for applications such as TCM knowledge base construction and automated spectrum–structure identification, thereby promoting the advancement of TCM informatics and data-driven research.