This project is about the modelling and optimization of HPLC mobile phases. We applied the methodology to develop a method for the screening of several pharmaceuticals. The project consisted on (a) finding the right equations for the description of the retention of solutes in RPLC, (b) using the fitted equations to predict retention in conditions different than the simmulated ones and (c) using some smart methodology to use these predictions to optimize a separation problem. In principle, there is nothing new in all these three topics: as for (a) it is known that that the relation of the logarithm of the capacity factor with the concentration of organic modifier is parabolic in RPLC (and it can be considered linear if we tolerate a certain ammount of error). Also, there is also nothing new in (b) and (c): these are steps in standard optimization strategies.
However, several aspects become interesting when one starts to look to the problem with more detail. For exmaple, suppose that we start to consider not only the predictions from the model in the optimization strategy, but also the errors introduced in the modelling. If we don’t consider that, our optimization strategy can be finding pointing to an optimal condition which in fact is quite uncertain since the error in the model at this condition is too large. This problem was discussed in detail in [12a]. Another improvement involved the use of internal standards to enhance the accuracy of the predictions [11a]. In another separated (but related) work we covered the application to the optimization mobile-phase gradients [7a]. In a later stage we developed a more general strategy involving the optimization of multi-linear gradients [15a, 20a, 21a]. For this last aspect of the project, it was important to check the error propagation of models when there is a transference of data between elution regimes (i.e. using information from isocratic elution to to predict gradients and viceversa). More details about this last piece of work involving error analysis can be found here.
This project was developed at the University of Valencia (Spain). Several people were involved. See authors of the publications for more details about authorship.