Barriers to Fully Automated Method Development using Machine Learning in Chromatography and the Role of Intuition

At Pittcon 2026 in San Antonio, Texas, USA, LCGC International spoke with 2026 LCGC International Emerging Leader Award Winner_—Bob Pirok from the University of Amsterdam, Amsterdam, The Netherlands— about his presentation: Machine Learning Algorithms in Automation for LC–MS Method: An Update on Bayesian Optimization Kernels.¹

Pirok’s research focuses on applying chemometrics and machine learning (ML) to analytical chemistry, particularly in multidimensional chromatography and automated method development.^2,3,4,5,6

His lecture explored how Bayesian optimization (BO) algorithms can automate LC–MS method development.¹In these systems, the algorithm iteratively selects experimental conditions, analyzes the resulting chromatographic performance, and learns how to propose improved conditions in a closed-loop workflow. Platforms such as AutoLC have demonstrated that fully unsupervised method development is feasible.

In this video interview, Pirok responds to the following questions:

• From your perspective, what are the biggest practical barriers to fully automated method development in real laboratories: instrumentation, algorithms, or the way we currently think about separations?

• Another challenge we see in automated method development is that models often do not transfer well between columns, instruments, or labs. How important is transferability for the future of AI in chromatography?

• Experienced chromatographers often have a strong intuition about which gradients or solvent combinations might work. Do you think this intuition can realistically be captured in machine-learning algorithms?

References

1. Pittcon LCGC International Lifetime Achievement and Emerging Leader Award Session: https://app.swapcard.com/event/pittcon-2026/planning/UGxhbm5pbmdfNDMxODgzNg==
2. Bos, T. S.; Boelrijk, J.; Molenaar, S. R. A.; et al. Chemometric Strategies for Fully Automated Interpretive Method Development in Liquid Chromatography. Anal Chem 2022, 94 (46), 16060–16068. DOI: 10.1021/acs.analchem.2c03169
3. Boelrijk, J.; Ensing, B.; Forré, P.; Pirok, B. W. J.
   Closed-Loop Automatic Gradient Design for Liquid Chromatography Using Bayesian Optimization. Anal Chim Acta 2023, 1242, 340789.
   DOI: 10.1016/j.aca.2022.340789
4. Molenaar, S. R. A.; Bos, T. S.; et al. Computer-Driven Optimization of Complex Gradients in Comprehensive Two-Dimensional Liquid Chromatography. J Chromatogr A 2023, 1707, 464306. DOI: 10.1016/j.chroma.2023.464306
5. Boelrijk, J.; Molenaar, S. R. A.; Bos, T. S.; et al. Enhancing LC×LC Separations through Multi-Task Bayesian Optimization. J Chromatogr A 2024, 1726, 464941. DOI: 10.1016/j.chroma.2024.464941
6. Bos, T. S.; Desport, J. S.; Buijtenhuijs, A.; et al. Composition Mapping of Highly Substituted Cellulose-Ether Monomers by LC–MS and Probability-Based Data Deconvolution. J Chromatogr A 2023, 1689, 463758. DOI: 10.1016/j.chroma.2022.463758

Biography
Bob W.J. Pirok obtained his PhD in analytical chemistry at the University of Amsterdam in 2019, defending his thesis cum laude. He is currently Associate Professor in Analytical Chemistry at the University of Amsterdam and Visiting Research Professor at Gustavus Adolphus College (USA). His research focuses on applying chemometrics and machine learning to automate and optimize chromatography–mass spectrometry systems, with emphasis on self-driving method development in one- and two-dimensional liquid chromatography