Data-driven development of liquid chromatography-mass spectrometry methods for combined sample matrices.

Area of Science:

• Analytical Chemistry
• Pharmacognosy
• Computational Chemistry

Background:

• Herbal medicine formulas (HMFs) are complex mixtures requiring sophisticated analytical techniques.
• Optimizing separation conditions for Liquid Chromatography-Mass Spectrometry (LC-MS) in HMF analysis is challenging due to numerous analytes and interfering compounds.
• Traditional optimization methods for LC-MS separation can be time-consuming and experiment-intensive.

Purpose of the Study:

• To develop a systematic, data-driven approach for optimizing LC-MS separation conditions for HMFs.
• To reduce the complexity and experimental workload associated with developing LC-MS methods for HMFs.
• To enable rapid and efficient method development for analyzing diverse herbal medicine combinations.

Main Methods:

• Extracted chromatographic peak parameters (e.g., retention times) from an LC-MS database of individual herbal medicines.
• Utilized machine learning to build predictive models correlating chromatographic and separation parameters (r > 0.996).
• Optimized separation parameters for HMFs based on model predictions, minimizing initial experiments.

Main Results:

• Successfully predicted chromatographic behavior of analytes and interfering compounds in HMFs using machine learning models.
• Achieved well-separated analytes in validation experiments for six different HMFs without prior HMF-specific experiments.
• Demonstrated the effectiveness of the data-driven approach in systematic and rapid LC-MS method development.

Conclusions:

• The proposed data-driven approach offers a systematic and efficient strategy for developing LC-MS methods for complex HMFs.
• This methodology significantly reduces the need for extensive experimental optimization, accelerating analytical method development.
• The approach allows for flexible adjustment of separation conditions to accommodate different analytes and HMF compositions.