Machine learning analysis of pharmaceutical cocrystals solubility parameters in enhancing the drug properties for advanced pharmaceutical manufacturing

Abstract

A new computational framework based on machine learning was developed for prediction of Hansen solubility parameters in preparation of pharmaceutical cocrystals with improved properties. The models of Kernel Ridge Regression (KRR), Multi-Linear Regression (MLR), and Orthogonal Matching Pursuit (OMP) were optimized in prediction of three Hansen solubility parameters. Each model’s performance was assessed via measuring Root Mean Square Error (RMSE), R2, Mean Absolute Error (MAE), and Monte Carlo Cross-Validation (CV) scores using a Tabu Search method for optimization. The results demonstrated that KRR outperformed other models for predicting solubility parameters in the formulation. This comparative evaluation offers valuable perspectives on selecting models for similar regression assignments, stressing the significance of choosing the right algorithm according to particular output demands. The results are useful for design of medicines and screening coformers with solubility enhancement in pharmaceutical co-crystallization.