Artificial intelligence for method development in liquid chromatography

Abstract

Method development in liquid chromatography is an important process in building up qualitative analytical methods that allow the separation and quantification of all compounds in a mixture. It is often a demanding process due to its time-consuming, resource intensive, and costly nature. This review explores the integration of artificial intelligence and machine learning to assist in and speed-up the method development process. The utility of Quantitative Structure-Retention Relation models in the screening phase of the method development is first addressed, with a particular focus on advanced molecular representations that utilize deep learning architectures, enabling more detailed molecular descriptions. Secondly, optimization algorithms that can automate and accelerate the optimization phase of the method development are discussed. Notable advancements include Bayesian optimization and reinforcement learning for the self-optimization of chromatographic parameters. Furthermore, artificial intelligence-based signal processing methods are reviewed, along with their role in the automation of the method development process. Despite these advancements, challenges remain in achieving a fully automated and experimentally efficient method development, and further improvements in molecular modelling, experimental design, and signal processing are needed. This review provides insights into current methodologies, future directions, and existing gaps in artificial intelligence-assisted method development, highlighting its potential impact in analytical chemistry.