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

Area of Science:

• Crystallization science
• Computational chemistry
• Pharmaceutical solid-state chemistry

Background:

• Solvate formation is crucial for pharmaceutical development, impacting drug stability, solubility, and bioavailability.
• Predictive models are needed to efficiently screen solvents for desired solvate formation.
• Curcumin (CUR) and its derivatives (demethoxycurcumin [DMC], bisdemethoxycurcumin [BDMC]) serve as relevant models due to their pharmaceutical interest.

Purpose of the Study:

• To compare statistical and thermodynamic methods for predicting solvate formation.
• To evaluate the performance of Statistical Frequency of Interaction for Multicomponent Prediction (SFIMP) and Conductor-like Screening Model for Realistic Solvents (COSMO-RS).
• To identify optimal predictors for guiding solvent selection in crystallization screening.

Main Methods:

• Utilized SFIMP and COSMO-RS computational methods to predict solvent interactions.
• Conducted comprehensive crystallization screening experiments to identify novel solvated and hydrated forms.
• Assessed individual predictor performance, focusing on hydrogen bond propensity (HBP).

Main Results:

• Several new solvated and hydrated forms of CUR, DMC, and BDMC were successfully obtained through crystallization screening.
• Hydrogen bond propensity (HBP) demonstrated the highest predictive performance among individual parameters.
• The combination of COSMO-RS with HBP provided the most accurate predictions for solvate formation.

Conclusions:

• The synergistic approach of combining COSMO-RS with HBP significantly enhances the accuracy of solvate formation prediction.
• These findings offer valuable insights for the rational design and efficient screening of multicomponent solid forms in drug development.
• This predictive framework can accelerate the discovery of optimal solid forms for pharmaceutical applications.