February 26–March 1, 2007, PITTCON® 2007, Chicago, IL

24-Hour Creation of Chromatographic Prediction Systems

Mike McBrien, David S. Bell, Alexey Galin, Vitaly Lashin

Abstract

For more than 20 years, there have been efforts to create a system to predict the elution of compounds based on their chemical structure. Chromatographic systems have become more robust, and computers have become faster, so that even gradient and mixed-mode retention systems can be modeled. These predictions of elution data for new compounds can be applied to the selection between generic methods for preparative scale work or purity estimation, starting points for method development, or even retention time-based confirmation of chemical structures. One of the requirements for the creation of a prediction system is the collection of elution data for a sufficient number and diversity of relevant compounds for each targeted chromatographic method. While the ideal approach is to collect this data during the normal course of experimentation in order to generate a knowledgebase without the need for extra measurements, this can become a barrier to the adoption of a system of prediction; for some laboratories, the collection of sufficient data can take months or longer. With the recent development of chemometric tools for the extraction of peak data from LC/MS datasets, there is a new opportunity to perform efficient extraction of hundreds of sets of retention data in a limited number of chromatographic injections, and with practically no manual evaluation of data. The approach involves the injection of extremely complex samples that are then deconvolved using chemometrics. It is now possible to collect elution data for a very large number of compounds and build a prediction model for a set of orthogonal chromatographic methods in less than a single day of combined experimentation and computation, with practically no user involvement. Practical application of this work will be discussed, as well as the design of compound training sets and performance of the resulting prediction algorithms.

Download the presentation in MS PowerPoint (1.05 Mb ZIP file) or Adobe Acrobat format (996 Kb PDF file).

Relevant Products: ACD/ChromGenius, ACD/Method Development Suite, ACD/MS Manager, ACD/IntelliXtract

Relevant Solution: Method Development