Abstract
•Computational analysis of drug solubility in supercritical CO2.•Investigations on the influence of pressure and temperature on solubility.•Validating the proposed models using measured dataset.
Preparation of drug nanoparticles has been studied and evaluated in this study based on supercritical-based processing as green technology. Computational works have been conducted to evaluate the possibility of manufacturing nanomedicine using this novel technology, and the results are compared with experimental measurements. Chlorothiazide, used as a diuretic and as an antihypertensive was considered as model drug in this work. For the modeling, we used a small data set consisting of two input features, namely temperature and pressure, and one output, namely solubility, in order to analyze the data. Tree ensemble models, including bagging and boosting based on decision trees, have been selected to analyze and model the data. Extremely randomized Trees (Extra Tree), Adaptive Boosting (AdaBoost), and Gradient Boosting models are specifically chosen for this modeling. The hyperparameters of the models were optimized with the help of genetic algorithm (GA) and finally the optimal models were obtained for each of the three methods. Finally, the models were evaluated with different methods. Based on the evaluations, the gradient boosting model showed the best results, and its score was 0.9820 with the coefficient of determination (R2-score) criterion. Also, the error of the final model with the MEA criterion is 1.51 × 10-2, with the RMSE criterion equal to 2.51 × 10-2, and the MAPE error value is 1.59 × 10-2.