Hansen Solubility Parameters in Pharmaceutical Cocrystal Screening: From Group Contribution Methods to Machine Learning-Assisted Prediction

Abstract

This review summarizes the theoretical basis of HSPs and discusses the major group contribution methods used for HSP estimation, including the Fedors, Hoy, and Van Krevelen approaches. The application of HSPs in pharmaceutical cocrystal screening, acceptance criteria for cocrystal prediction, and representative case studies involving poorly soluble drugs are critically examined. The advantages and limitations of HSP-based prediction methods are also discussed, highlighting the need for experimental validation of predicted cocrystal systems. Furthermore, recent advances involving machine learning, artificial intelligence, molecular descriptors, and COSMO-RS-assisted screening strategies are reviewed as emerging approaches to improve prediction accuracy and coformer selection. The integration of traditional group contribution methods with modern computational technologies is expected to accelerate pharmaceutical cocrystal discovery and reduce experimental workload. Overall, HSP-based screening remains a valuable and practical tool for early-stage cocrystal development and continues to evolve through advances in data-driven and computational methodologies.