Analysis of nanonization and purification of organic compounds via green supercritical processing: Model development using advanced hybrid techniques

Abstract

Complex relationship between process parameters and critical quality attributes in nanonization of drugs based on supercritical processing demands for developing advanced hybrid computational models. In this research, the complicated relationships between temperature, pressure, solvent density, and solubility of Nystatin in supercritical carbon dioxide for green processing of nanoparticles was explored via advanced hybrid models. For the first time, three base regression models: K-Nearest Neighbors (KNN), Bayesian Ridge Regression (BRR), and Polynomial Regression (PR) in combination with Political Optimizer (PO) for correlation of solubility of organic compound as well as solvent density was developed. Additionally, an ensemble approach, Bagging, is employed to enhance predictive performance of the models. Bagging ensemble technique combined with BRR predicts Nystatin solubility with an R2 of 0.89459. Bagging combined with KNN produces a score of 0.9162, showcasing its strong predictive capabilities. Moreover, bagging coupled with PR exhibits exceptional predictive performance, attaining a score of 0.99819. When predicting solvent density, bagging in conjunction with BRR yields an R2 value of 0.81069, while bagging paired with KNN achieves a higher score of 0.90542, demonstrating its proficiency in modeling solvent density. Bagging with PR surpasses the other models, attaining a significant R2 of 0.99035, highlighting its superior predictive accuracy.

Keywords

• Nanoparticles
• Green processing
• Modeling
• Machine learning
• Bagging