Integrating GC-MS profiling and portable mass spectrometry with machine learning for comprehensive chemical characterization and rapid authentication of nine Curcuma species.

Plants of the genus Curcuma are vital medicinal resources; however, their highly similar chemical profiles and morphological features present substantial challenges for accurate species authentication. Here, we established a comprehensive analytical strategy for the precise differentiation of eight medicinal Curcuma species and one counterfeit by integrating untargeted gas chromatography-mass spectrometry (GC-MS) profiling with rapid in situ portable mass spectrometry (PMS) and machine learning. GC-MS analysis tentatively identified ten predominant volatile components, whose potential biological targets and signaling pathways were elucidated via network pharmacology. Chemometric analysis of GC-MS metabolic profiles further enabled the screening of core differential markers driving species discrimination. For rapid on-site detection, in situ PMS fingerprints were acquired and processed using a characteristic ion-based binarization strategy following base peak normalization. When coupled with advanced machine learning algorithms, particularly Ensemble and Efficient Linear classifiers, the system achieved 100% classification accuracy with high computational efficiency. Together, this dual-platform approach provides an effective, and broadly applicable method for quality control and rapid authentication of multi-origin traditional medicines.