Environmental suitability-driven metabolomics and machine learning study on origin variation and traceability of Gastrodia elata Blume

The study established a species distribution model using Maximum Entropy (MaxEnt) and analyzed untargeted metabolomics data of phenolic compounds in GEB to assess the differences in metabolites within its suitable habitats of Gastrodia elata Blume (GEB). It also integrated machine learning with Attenuated Total Reflectance-Fourier Transform Infrared Spectroscopy (ATR-FTIR) to develop a traceability model for its geographical origin. The results indicate that under current climatic conditions, the primary suitable habitats for GEB are concentrated in southwestern China. The key environmental factors affecting its growth suitability are bio12, bio09, and bio02, with an optimal elevation range of 197 m to 3,756 m. The content of gastrodin and parishins (A, B, C) was predominant in the ZT region, while p-hydroxybenzyl alcohol, parishin E, and total parishins were notably higher in the DF region. Metabolites such as [6]-shogaol, 1-O-sinapoyl-beta-D-glucose, and erianin may serve as marker compounds distinguishing ZT from other regions. The DD-SIMCA model achieved perfect classification performance in distinguishing ZT from other origins with a PC count of 4. Additionally, the ResNet model, constructed based on ATR-FTIR synchronous 2DCOS images, achieved 100% classification accuracy for GEB from multiple origins, promising applications for origin traceability of GEB.