< Back to previous page

Project

Developing machine learning algorithms for chromatographic method development

Chromatography is a laboratory technique to separate a mixture into its components. Unraveling the exact composition of a mixture is of great importance in a large number of disciplines, including the pharmaceutical industry, chemical industry, environmental analysis etc. The ultimate goal of this PhD project was to develop machine learning (ML) algorithms as tools to improve the separation of mixtures into their components, as well as identifying these components, by focusing and improving on different submodules in the chromatographic workflow. Nowadays, the process of successfully separating a mixture is often still based on a trial-and-error approach, which can be very laborious and costly. By implementing ML algorithms, it is hoped that both the rate of successful separations and number of steps to attain a successful separation, can be reduced. Specifically, ML algorithms were developed for (1) signal processing, both to denoise and perform peak detection on the output chromatograms; (2) finding the optimal set of scouting runs to be performed, to obtain optimal retention models; (3) identifying components (underlying molecules) based on a link between molecular structures and retention times (also known as quantitative structure-retention relationships). Importantly, improving on these three steps would result in (1) more components being detected, (2) more accurate retention time predictions of future chromatographic runs (consequently reducing the number of tries needed to attain a separated mixture), and (3) better identification of the components in the mixture.

Date:31 Jul 2018 →  26 Oct 2022
Keywords:Machine learning, Chromatography, Pharmaceutical analysis, Artificial intelligence
Disciplines:Analytical chemistry, Drug discovery and development, Pharmaceutics, Machine learning and decision making, Cheminformatics, Chemometrics
Project type:PhD project