Advanced modeling and intelligence-based evaluation of pharmaceutical nanoparticle preparation using green supercritical processing: Theoretical assessment of solubility

Abstract

Modeling and simulations based on machine learning techniques were conducted in this research for determination of pharmaceutical solubility in supercritical solvent for the sake of green processing development in the area of pharmaceutics. By this method, drug nanoparticles can be formed and controlled for better solubility and consequently bioavailability. We modeled the solubility of the drug namely desoxycorticosterone acetate (DA). The temperature and pressure are two inputs for the developed models, whereas the target output is the solubility of DA in the solvent. Indeed, models with two inputs and one output are developed in this study. This dataset contains 45 rows of data collected at five distinct temperatures and nine pressure levels. Three novel models are employed here including: HHO-MLP, HHO-RR, and HHO–KNN for theoretical determination of the drug solubility in the solvent. They are in fact MLP, Ridge Regression, and KNN models coupled with Harris Hawks optimizer (HHO) for hyper-parameter optimization. The HHO-MLP model has the best performance among models with R2 of 0.995, RMSE error rate of 0.194 and maximum error equal to 0.327. The model's findings can be used for estimation of the solubility data in wide range of T and P.

Keywords

• Pharmaceutics
• Green chemistry
• Multilayer perceptron
• Machine learning
• Pressurized solvent