Multidimensional chromatographic fingerprint fusion with machine learning: Entropy-based feature evaluation for TCM...

Area of Science:

• Analytical Chemistry
• Cheminformatics
• Pharmacognosy

Background:

• Chromatography is vital for traditional Chinese medicine (TCM) quality control.
• Complex chromatographic data with overlapping signals hinders accurate feature identification.
• Existing methods struggle with the intricate chemical profiles of TCMs.

Purpose of the Study:

• To develop an integrated approach combining multidimensional chromatographic fingerprinting and machine learning.
• To trace the molecular origins of characteristic compounds in Fufang E'jiao Jiang (FEJ).
• To establish a scalable computational framework for quality control in complex samples.

Main Methods:

• Multidimensional chromatographic fingerprinting (TLC, LC-HRMS) to generate datasets (>1700 features).
• Machine learning algorithms (e.g., random forest) for discriminative feature selection.
• Modified entropy and obstacle factors for feature evaluation and weighting.

Main Results:

• Identified 5 patterns in FEJ and 7 in intermediates, primarily ginsenosides.
• Lobetyolin and ginsenoside Rf were identified as key quality-related markers.
• Experimental verification confirmed effective deconvolution of overlapping signals.

Conclusions:

• The integrated approach efficiently identifies key quality markers in complex chromatographic fingerprints.
• The method offers a general, scalable computational framework applicable to various complex samples.
• This strategy enhances the quality control of TCM and other complex chemical systems.