Using Chromatographic Knowledge to Construct Landing Platforms for Efficient Method-development Strategies

Strategies for automating method development and optimization in liquid chromatography are seeing a spurt in efficiency, thanks to the great current interest in machine-learning and artificial-intelligence techniques. The notorious bottleneck of systematic method-development method, the significant time required for each chromatographic experiment, can be largely overcome by maximizing the information gained from each experiment, and to augment the experimental data with high-quality simulated data. With some imagination, method-development strategies have been taken from bicycle speed to car speed by building retention models for each analyte (“interpretive methods”), while hybrid strategies, such as Bayesian optimization, take us to the airplane level.

What we need now is good airports, for both the take off and the landing of the optimization strategies. At either end, we need to make the best possible use of our chromatographic knowledge. We will have very limited success if we fail to select good starting conditions, or if we direct our process at a sub-optimal target. There is much room for chromatographers to improve the results obtained using contemporary artificial-intelligence methods by anchoring them in chromatographic knowledge.