Abstract
Background: Solubility of drug/drug-like molecules plays a major role in pharmaceutical sciences for obtaining suitable solvent system for the desired pharmacological response. Experimental measurement is time-consuming and costly, therefore, developing a computational procedure to predict the solubility of drugs in different mono-solvents and temperatures is necessary. No accurate ab initio prediction method is available so far, and as an alternative, one may use empirical/semi-empirical models trained by using a single experimental data point.
Methods: To achieve this goal, the available solubility data sets were collected from the recently published articles and selected a single data point of each dataset at 298.15 K to train two models adopted from the Hildebrand solubility approach which proposed previously by our research group. After obtaining two models’ parameters, the rest of solubility data points in datasets were predicted. The accuracy of models was evaluated by computing the mean percentage deviation (MPD) of the predicted data.
Results: The low value of overall MPDs (≤ 19.5%) obtained revealed that the models could be employed as a practical strategy for the prediction of drugs solubility in mono-solvents at different temperatures with an acceptable prediction error.
Conclusion: The proposed computational method could be successfully applied in the pharmaceutical industry where solubilization of drugs is highly in demand.