Advances in the application of artificial intelligence in mass spectrometry-based analysis of traditional Chinese medicine: compound identification and metabolic pathway elucidation.

Traditional Chinese medicine (TCM), characterized by its multi-component, multi-target, and multi-pathway nature, presents considerable challenges in the identification of chemical constituents and elucidation of metabolic mechanisms. TCM samples encompass a wide range of materials, including crude herbal parts, processed products, in vitro cell cultures, and in vivo biological specimens, each contributing to the complexity of analysis. MS, as a pivotal analytical tool for uncovering the material basis of TCM, has been widely employed for compound identification and in vivo metabolic pathway analysis, owing to its high throughput, sensitivity, and resolution. However, the inherently high-dimensional, noisy, and complex nature of MS data poses significant limitations to traditional analytical methods in terms of data processing efficiency and structural identification accuracy. In recent years, artificial intelligence (AI) technologies, particularly machine learning (ML), and deep learning (DL) models, have demonstrated remarkable potential in spectral interpretation, structure prediction, and metabolic pathway modeling within the context of MS-based TCM research. This review systematically summarizes the latest advances in the application of AI in TCM MS analysis, with a particular focus on two key areas: the utilization of AI for rapid qualitative analysis of complex TCM compounds, including spectral preprocessing, feature extraction, structural attribution, and isomer differentiation; and the role of AI in metabolite identification and reconstruction of in vivo metabolic pathways (Alvarado, Sci Eng Ethics, 29(5):32, 2023), encompassing metabolite screening, network modeling, and multi-omics integration. Furthermore, we critically discuss current challenges impeding further progress, such as the lack of high-quality MS databases, limited interpretability of AI models, and insufficient capabilities for cross-modal data fusion. Finally, we propose future directions for the field, emphasizing the importance of building interpretable, generalizable, and integrative AI frameworks. In summary, the convergence of AI and MS technologies is reshaping the paradigm of TCM research from empirical investigation to data-driven intelligence, thereby opening new avenues for the modernization of TCM and precision pharmacological studies.