Abstract Title: | Towards Multifactorial Method Development via Predictive Elution Window Stretching and Shifting |
Abstract Type: | Poster |
Session Choice: | Big Data Chemometrics and Method Development(In-Silico)(KVCV) |
Presenter Name: | Ms Gitte Coopmans |
Co-authors: | Prof Gert Desmet Prof Sebastiaan Eeltink |
Company/Organisation: | VUB |
Country: | Belgium |
Abstract Information :
The selection of the chromatographic separation conditions leading to the best and fastest separation of a given sample is a very complex problem. Whereas this problem is in many cases still solved via time-consuming trial-and-error approaches, there is a growing interest in using Computer-Aided approaches. Recently, our group has contributed to this work by proposing the so-called predictive elution window stretching and shifting (PEWS2) method. In this approach, the analyst is automatically guided to explore the most promising areas of the parameter space by directing the first and the last eluting compound to some given pre-set values.
Up till now, the PEWS2-method only optimised the gradient parameters. In the present work, the method was extended to account also for the effect of pH and temperature on the chromatographic separation selectivity. This was achieved via a factorial model composed of a non-linear Neue Kuss factor to describe the mobile phase composition effect, a van't Hoff factor to describe the enthalpic retention effect and a sigmoidal factor to describe the pH-effect. Two search methods (for-loop search and grid search) were developed and compared. The grid search proved to be about an order of magnitude faster, but is very demanding in terms of Random Access Memory. The search methods could be significantly accelerated by splitting the parameter set in two subgroups (so-called 'divided search').
An algorithm was developed to steer the elution of two components to a given pre-set value of the retention time. To direct the elution behaviour of a mixture, the concept of pseudocompounds was introduced to accommodate the fact that the first and/or last eluting compound of a mixture not always relate to the same analyte. Applying the steering algorithm on pseudocompounds allowed to cover the entire space of physically achievable conditions of a mixture.
It was also possible to apply the newly developed extended PEWS2-method to a real mixture, composed of alkylphenones and organic acids. In a first round, a series of scouting runs was performed at different pH, temperature and gradient program conditions to determine the retention parameters of a pseudocompound representing the first and last eluting analyte over the entire range of possible separation conditions. Subsequently, the retention parameters were used to predict a number of conditions wherein the first and last eluting analyte of the mixture are spread over a sufficiently wide range and with a large variability in start and ending points. Finally, the proposed conditions were run and the new conditions indeed yielded a number of very high separation resolution conditions.