Comprehensive Discrimination of Amomi Fructus From Different Origins Using UHPLC‐Q‐Orbitrap MS, HS–GC–MS/MS, NMR and MIR Technologies Based On Data Fusion Strategies

TL;DR

This study uses advanced analytical techniques and data fusion strategies to accurately identify the origin of Amomi Fructus, an important spice and traditional Chinese medicine.

Contribution

The study introduces a mid-level data fusion model combining random forest algorithms for improved classification of Amomi Fructus origins.

Findings

A mid-level data fusion model using random forest algorithms achieved the best classification of Amomi Fructus origins.

27 differential compounds were identified and verified to significantly improve classification accuracy.

The combined use of multiple analytical techniques provided a more comprehensive and accurate analysis method.

Abstract

Amomi Fructus (SR) is an important edible herb widely used as a spice and traditional Chinese medicine. To comprehensively solve the serious practical problems of origins and species confusion in SR, the systematic characterization methods were established by liquid chromatography–mass spectrometer, gas chromatography–mass spectrometer, nuclear magnetic resonance and infrared spectroscopy. A total of 286 compounds and functional group information were detected. The classification of SR from different origins was performed by data fusion models built using random forest (RF) and other algorithms. A mid‐level data fusion model (an RF model established after combining the features selected by RF and RF–RF) performed the best classification. Then 27 differential compounds (including flavonoids, polyphenols and terpenoids) and their functional group information were screened for external verification and could significantly improve the groups’ separation effect just by simple principal component analysis. A more comprehensive and accurate means of analysis was found.

Figures