Neural simulation and experimental investigation of Chloroquine solubility in supercritical solvent

Understanding drug solubility in various solvents is of great importance to the pharmaceutical industry for design of the process. To date, various modeling approaches including Equation of State (EoS) and Semi-empirical correlations have been developed for drug solubility prediction. A neural-based modeling method has been proposed in this work for prediction of drug solubility of solvents in condense state. A supercritical solvent has been selected as the condense solvent due to its properties. The modeling approach is based on artificial intelligence, which has been proposed and validated by comparison with experimental data. The modeling results have been compared with measured data, and agreement was observed. Chloroquine was selected as the APT in this work due to its extensive usage for treating malaria and Coronavirus. The solubility has been measured experimentally at different temperatures and pressures. 32 experiments were carried out and the solubility was obtained using a mole fraction unit. The neural network was developed considering temperature and pressure as the inputs, while the solubility was set as the output. The results of training and validation indicated that the developed neural network is robust in predicting solubility, and high accuracy with R of more than 0.99 was obtained for both the training and validation set. Finally, it was indicated that pressure has significant effect on the drug solubility, and a direct relationship was obtained between solubility and pressure.