Unraveling the spatial chemical heterogeneity of Lentinula edodes: Integrative metabolomics and machine learning for cultivar and origin classification.

As a major edible mushroom, Lentinula edodes requires accurate cultivar and origin authentication. Conventional homogenate-based metabolomics obscures spatial chemical information. This study establishes an integrated framework combining UHPLC-Orbitrap MS metabolomics, DESI-MSI imaging, and machine-learning modeling to decode the spatial chemical heterogeneity of three representative cultivars (Q2, K2, S1). Untargeted LC-MS profiling identified 101 differential metabolites across key amino acid, organic acid, nucleoside, and phenolic pathways. DESI-MSI visualized 65 of them, revealing tissue-specific chemical distributions inaccessible to traditional strategies. By integrating abundance and spatial intensity, feature selection using RF, PLS-DA, LASSO, and RFE yielded a robust seven-metabolite biomarker panel (including d-mannitol, L-malic acid, riboflavin, and p-coumaric acid, etc.). These chemical-spatial signatures formed distinctive "Integrated Fingerprint Cards" for each cultivar, enabling reliable classification and geographical traceability. This chemical-spatial paradigm provides a powerful foundation for future real-time, in situ authentication of high-value agricultural products.