Determination of paclitaxel anticancer drug solubility in supercritical CO2: Thermodynamics modeling and machine learning approach

چکیده مقاله

To facilitate the effective design of supercritical fluid (SCF) processes aimed at micro- or nanosizing solid pharmaceuticals, obtaining solubility data in environmentally friendly solvents such as pressurized carbon dioxide (CO₂) is essential. Solubility assessment represents a critical first step in evaluating SCF technologies. This study introduces a statistical methodology to experimentally determine the solubility of paclitaxel (Pac) in supercritical CO2. UV-vis spectrophotometric studies were carried out under pressures ranging from 120 to 270 bar and temperatures ranging from 308 to 338 K. Three distinct modeling approaches were used to predict and correlate the experimentally determined solubility of paclitaxel: (i) a collection of six density-based empirical models; (ii) a hybrid of the Peng-Robinson (PR) equation of state and the van der Waals quadratic mixing rule; and (iii) machine learning procedures, including fifteen non-linear regressions. A solubility range of 0.0017 to 0.077 g/L was observed for paclitaxel. At a steady temperature, the paclitaxel mole fraction increased as the pressure rose, albeit a crossover occurrence was noted. While all methods achieved adequate levels of correlation accuracy, the Méndez-Santiago & Teja (MT) model outperformed the others in terms of predictive power, with an AARD of only 4.06%. For the first time semi-empirical correlations were used to estimate the paclitaxel/Sc-CO₂ system enthalpies as 〖∆H〗_"tot" = 28.04 kJ/mol, 〖∆H〗_"sol" = -19.11 kJ/mol, and, 〖∆H〗_"vap" = 47.15 kJ/mol.