Contribution To Modeling The Solubility Of Solid Drugs In Different Solvents

Abstract

The primary purpose of this thesis is to contribute to the area of solubility prediction with three advanced modelling techniques built with the use of powerful input parameters based on molecular information; Addresses the prediction and correlation of the solubility of solid drugs in two different green solvents, namely: supercritical carbon dioxide and pure ionic liquids. This thesis covers a variety of approaches to provide an account of the current state of the art and justifies the choices of the proposed methods for solubility prediction. The first contribution introduces the use of Quantitative Structure–Property Relationship techniques with an Artificial neural networks algorithm for correlating the solubility of 148 drug compounds in supercritical carbon dioxide. The second contribution deals with a comparison study between seven selected meta-heuristic algorithms: Dragonfly optimizer, Ant Lion optimizer, Grey Wolf optimizer, Artificial Bee Colony optimizer, Particle Swarm optimizer, Whale optimizer, and a hybrid Particle Swarm with Grey Wolf optimizer in fine-tuning hyperparameters of support vector regression algorithm on a database of 168 pharmaceutical compounds solubilities in supercritical carbon dioxide. The third contribution shows the powerful combination of Quantitative Structure–Property Relationship techniques with a conductor-like screening model for Real Solvents based techniques in modelling 14 active pharmaceutical ingredients solubilities in various ionic liquids. The Good-of-fit, robustness, and predictive power of all proposed models were confirmed with various statistical and graphical metrics