Comparative Assessment of Statistical and Thermodynamic Prediction Methods for Solvate Formation: A Case Study with Curcumin and Its Derivatives

This study compares statistical and thermodynamic methodologies for predicting solvate formation using curcumin (CUR) and its derivatives demethoxycurcumin (DMC) and bisdemethoxycurcumin (BDMC) as models. We evaluated the performance of Statistical Frequency of Interaction for Multicomponent Prediction (SFIMP) and Conductor-like Screening Model for Realistic Solvents (COSMO-RS) methods to identify solvents likely to form solvates. A comprehensive crystallization screen yielded several new solvated and hydrated forms. Our results show that hydrogen bond propensity (HBP) performed best among individual predictors, while COSMO-RS combined with HBP yielded superior predictive accuracy overall. These insights aid rational design and screening of multicomponent solid forms in pharmaceutical development.