Closing the loop: Faster, smarter, more automated chromatography
Sebastian Detlefsen, group head of chromatography and purification, and Jim Boelrijk, data scientist at Evonik, showcase a newly developed approach for automated liquid chromatography *
Liquid chromatography( LC) is one of the most widely used analytical techniques worldwide, both in terms of the number of practitioners and the financial investment involved. It plays a critical role across industries, from pharmaceuticals to chemicals, enabling precise separation and analysis of complex mixtures.
In recent years, machine learning has begun to transform LC method development, making it faster, smarter and less dependent on manual trialand-error. Among these approaches, Bayesian optimisation stands out. This data-driven technique learns from each experiment and predicts the most promising next step, dramatically accelerating optimisation. 1, 2
Gradient design is critical for achieving accurate, efficient and reproducible results with LC. Traditionally, designing gradients for complex mixtures has been a painstaking process requiring countless manual iterations.
Each step requires the interpretation of chromatograms, the adjustment of parameters and stepwise improvement of results, a cycle that can stretch over days or even weeks. With expert knowledge and development time often scarce, developing a fully optimised method that includes the design of a complex, multi-linear gradient is often not possible.
Evonik, in collaboration with Agilent, has broken this cycle with a fully automated, intelligent system. We have created an automatic gradient
Figure 1 – Workflow of BoFire
Figure 2- Example separation problem & result of optimisation
62 SPECIALITY CHEMICALS MAGAZINE ESTABLISHED 1981