Modelling of Carotenoids Solubility in Supercritical Carbon Dioxide Using Quantitative Structure-property Relationships

Prediction of solubility in SuperCritical CO2 (SC-CO2) as a function of system pressure (P) and temperature (T) aids selection of process condition for extraction processes. Several equations have being used to correlate solubility as a function of P and T, but best-fitting procedures typically demand a large set of experiments. Previously, other groups have developed semi-empirical models to predict solubility of different compounds in SC-CO2. Our objective is to develop a semi-empirical model to predict the solubility of carotenoids in SC-CO2 under different pressure and temperature conditions, using a small set of descriptors obtained from their equilibrated 3D structure. Experimental solubility of selected carotenoids in SC-CO2 at different pressure and temperature was used to build the model, using their solubility parameters according to Chrastil equation. Descriptors were calculated from the solute structures after molecular dynamic simulations in implicit CO2, from which carotenoids were separated in clusters. Descriptors were ranked, and Quantitative Structure-Property Relationships (QSPRs) were built using a small set of descriptors and a subset of the experimental values for solubility, using both linear regression and Artificial Neural Networks. Further experimental data is required to validate the model and to be able to predict outside of the training set, without the need to run a very large set of experiments.