Chemometric Strategies for Fully Automated Interpretive Method Development in Liquid Chromatography

Area of Science:

• Analytical Chemistry
• Chromatography

Background:

• Conventional liquid chromatography (LC) method development relies heavily on analyst expertise.
• Automated approaches are needed to improve efficiency and reproducibility in LC method development.

Purpose of the Study:

• To develop a novel, open-source algorithm for automated and interpretive LC method development.
• To create a closed-loop workflow for unsupervised, automated LC system interaction.

Main Methods:

• Developed an open-source algorithm, "AutoLC," for automated LC(-mass spectrometry) separations.
• Investigated challenges in peak tracking, retention modeling, gradient profile design, and chromatogram simulation.
• Tested two strategies: retention modeling and Bayesian optimization machine learning.

Main Results:

• The algorithm achieved optimal performance within 4-10 iterations using both tested strategies.
• Retention modeling proved more efficient but dependent on peak tracking.
• Bayesian optimization offered flexibility but faced scalability limitations.

Conclusions:

• The developed algorithm enables efficient, automated LC method development.
• The modular design ensures compatibility with existing and future LC data handling algorithms.
• AutoLC offers a powerful tool for optimizing LC separations based on resolution and analysis time.