Cocry-pred: A Dynamic Resource Propagation Method for Cocrystal Prediction.

Area of Science:

• Pharmaceutical Science
• Computational Chemistry
• Materials Science

Background:

• Drug cocrystallization enhances physicochemical properties without altering chemical structure.
• Identifying suitable coformers is a significant challenge in drug development, requiring extensive resources.
• Existing methods for coformer prediction are often resource-intensive and time-consuming.

Purpose of the Study:

• To develop a novel and efficient cocrystal prediction model, Cocry-pred.
• To streamline the identification of potential coformers for target drug molecules.
• To improve the efficiency and success rate of cocrystal screening and design.

Main Methods:

• Utilized the Network-Based Inference (NBI) algorithm for dynamic resource propagation.
• Incorporated topological data from cocrystal networks and molecular substructure information.
• Evaluated 13 molecular fingerprints and optimized hyperparameters (α, β, γ) for model performance.

Main Results:

• Cocry-pred achieved a high performance with an Area Under the Curve (AUC) of 0.885 and a R-squared (RS) of 0.108.
• The model successfully predicted potential coformers for Apatinib.
• Seven Apatinib cocrystals were synthesized, with two yielding single-crystal structures, validating the model's reliability.

Conclusions:

• Cocry-pred offers a powerful and efficient tool for cocrystal screening and design.
• The model significantly improves the process of identifying suitable coformers.
• This advancement provides valuable insights for accelerating drug property enhancement through cocrystallization.