Development of direct infusion quadrupole mass spectrometry database and intelligent identification system for 100...

Counterfeiting is widespread in the herbal medicine market, especially for roots and rhizomes. Current authentication methods for these herbs remain limited and subjective, posing challenges to quality assurance and authenticity verification.

To overcome the limitations of low efficiency and high subjectivity in traditional identification methods for root and rhizome traditional Chinese medicine (RRTCM), this study aims to develop a high-throughput intelligent identification system based on direct infusion QDa mass spectrometry (DI-QDa-MS) for the rapid and objective identification of 100 RRTCM, including 88 herbal medicines and 12 processed products.

MS data of 100 RRTCM were acquired using DI-QDa-MS. An automated data processing pipeline was developed via machine learning and cosine similarity algorithms, alongside the establishment of a standardized DI-QDa-MS database and feature ion library. This study innovatively developed three intelligent identification methods: k-nearest neighbor (KNN) model, MS similarity analysis, and binary coding. Their accuracy and computational efficiency were systematically validated using 105 batches of blind samples.

All three methods achieved validation success rates ≥95 %, with MS similarity analysis performing optimally. This established a standardized workflow for intelligent identification of 100 RRTCM based on MS similarity. Inter-day and intra-day precision values were both <3 %, confirming the reliability of the methods.

The identification technique proposed in this study, which leverages the cosine similarity algorithm, establishes a high-throughput approach for standardizing RRTCM quality. This method contributes to strengthening TCM market regulation and advancing TCM quality standards.