Role of Artificial Intelligence in Lc-Ms/Ms Metabolomics

Abstract

Artificial intelligence is transforming LC‑MS/MS metabolomics by automating data handling and improving metabolite identification, leading to higher sensitivity, reproducibility, and clinical relevance. Machine learning and deep learning enhance baseline correction, noise reduction, peak detection, deconvolution, and retention‑time alignment, reducing manual curation. Tools such as PeakBot, MS2DeepScore, MIST‑CF, SingleFrag, and CFM‑ID strengthen spectral matching, fragmentation prediction, and de novo annotation, expanding detectable metabolites. AI‑driven feature selection and classification support robust biomarker discovery, disease diagnosis, and multi‑omics integration for systems‑level pathway insights and precision medicine. Remaining challenges include limited annotated data, inter‑laboratory variability, model opacity, computational cost, and lack of standardized benchmarks. Future priorities are unified, explainable end‑to‑end AI workflows, shared LC‑MS/MS repositories with community benchmarks, and accessible cloud platforms to enable routine AI‑driven metabolomics in research and clinical practice.