Characterization of volatile organic compounds in different parts of Daphne tangutica by GC-IMS combined with machine learning.

Daphne tangutica Maxim. is a medicinally important plant whose optimal part selection and quality evaluation are critical for its rational exploitation and sustainable utilization. In this study, a rapid and reliable analytical strategy was developed to discriminate its root bark, stem bark, and leaves by integrating Gas Chromatography-Ion Mobility Spectrometry (GC-IMS) with machine learning. GC-IMS provides significant advantages for detecting volatile organic compounds (VOCs) through an orthogonal separation capability, sequentially resolving analytes by chromatographic retention and ion mobility differentiation to deliver high sensitivity and visual spectral fingerprints without extensive sample pretreatment. Following headspace incubation optimization, 75 VOCs, including 16 isomer groups, were identified. Chemometric analyses, including hierarchical cluster analysis (HCA) and principal component analysis (PCA), revealed distinct part-specific profiles. Notably, leaves exhibited distinct volatile fingerprints compared with traditional medicinal parts (root bark and stem bark), with stronger responses for several discriminating features. Orthogonal partial least squares discriminant analysis (OPLS-DA) characterized 24 differential markers (variable importance in projection, VIP > 1) driving part differentiation. To facilitate the efficient analysis of the high-dimensional and complex datasets, five machine learning models were constructed. The Deep Neural Network (DNN) and Random Forest (RF) models, both endowed with strong nonlinear feature extraction capabilities, achieved the highest classification accuracy and F1-score of 0.9881. Overall, integrating GC-IMS with machine learning constitutes a powerful and intelligent platform for revealing metabolic differences among plant parts, providing a robust technical framework for the comprehensive development and sustainable utilization of D. tangutica resources.