Determination of 5-fluorouracil anticancer drug solubility in supercritical CO2 using semi-empirical and machine learning models

چکیده مقاله

In order to provide the facilities to design the supercritical fluid (SCF) processes for micro or nanosizing of solid solute compounds such as drugs, it is essential to obtain their solubility in green solvents like pressurized CO2. This important role is the first stage for assessing each SCF technology. A statistical method was developed for the first time and employed to determine 5-fluorouracil (5-Fu) solubility. The measurements were performed at different pressures (120-270 bar) and temperatures (308-338 K) through UV-vis spectrophotometry, for the first time. The solubility was obtained between 0.0024 and 0.0176 g/L. The 5-Fu mole fraction at constant temperature, increases with an increase in pressure. Whereas, a crossover point has been seen. Three models with different approaches were applied to correlate and model the experimental data set: (i) seven density-based models, (ii) PR equations of state (vdW2 mixing rule), and (iii) machine learning-based models, namely non-linear regressions, Random Forest, Gradient Boosting, Decision Tree, and Kernel Ridge. All tested models successfully correlate and model the solubility data within an acceptable accuracy. Meanwhile, the empirical model suggested by Sodeifian model 2, is superior with the lowest AARD% (AARD=4.12%). Finally, to-tal, solvation, and vaporization enthalpies of the drug/Sc-CO2 binary system were determined using semi-empirical correlations, for the first time.