Predição da solubilidade de fármacos em sistemas cosolventes usando algoritmos de inteligência artificial

Introduction: Solubility is one of the most important physicochemical properties in the pharmaceutical sciences and is involved in processes ranging from drug development to environmental biotoxicity assessment. Accurate prediction of solubility, especially aqueous solubility, has been a scientific challenge due to the complexity of the dissolution process, which involves lattice energies, solvation, solute ionization, and solute-solvent interactions. While various approaches have been used to predict solubility, such as semi-empirical models and group contribution methods, accurate prediction remains difficult. Statistical and machine learning approaches often use numerous descriptors, making rational interpretation and refinement of prediction models difficult. Purpose: This study proposes an approach based on autonomous learning, specifically using a neural network model. Experimental and theoretical descriptors are carefully selected according to their relevance to physicochemical aspects of the dissolution process. Methodology: A solubility data set of four structurally related sulfonamides (sulfadiazine, sulfamerazine, sulfamethazine, sulfamethazine, and sulfacetamide) in different solvents under different temperature conditions and cosolvent composition is used to develop an artificial neural network model using less than 37 descriptors. Results: The developed AI algorithm shows acceptable correlation with experimental data, suggesting its potential as an approximation tool for optimizing solubility-related processes in the pharmaceutical industry.