Solubility enhancement of pantoprazole sodium sesquihydrate through supercritical solvent: Machine learning based study

Theoretical/experimental measurement and optimization of solubility is considered as an active field of research in different scientific areas such as pharmaceutics. Novel orally administered therapeutic medicines have recently suffered from low solubility and bioavailability. Therefore, finding better methods to improve the drug solubility and bioavailability are of prime importance. Application of CO₂ supercritical fluid (CO₂SCF) as a well-known procedure has attracted numerous attentions in pharmaceutical industry to increase the drug solubility and bioavailability due to its low flammability, reasonable cost, and negligible detrimental impacts for ecosystem. Pantoprazole sodium sesquihydrate is an extensively-used drug, which is indicated for the alleviation of the undesirable signs of esophagitis caused by stomach acid due to gastroesophageal reflux disease (GERD). The purpose of this research is to analyze and optimize the solubility of Pantoprazole sodium sesquihydrate using four different tree-based models where we have temperature and pressure as input features. These models include the regression tree, the boosting model based on it (gradient boosting), the bagging model based on it (random forest), and finally a more recently developed boosting model (extreme gradient boosting). All of these models, after optimization with their hyper-parameters, show an R²-score of more than 0.9. Finally, extreme gradient boosting was selected as the main model of this work, which has a 1.07 × 10^-2 maximum error and an RMSE error rate of 5.85 × 10^-3.