AI Enabled Pharmaceutical Cocrystal Screening

Solid form selection is consequential well before a drug reaches clinical trials. The wrong choice can affect bioavailability, stability, manufacturability, and regulatory standing. Cocrystal systems offer a meaningful pathway to optimize API physicochemical properties, but the combinatorial scope of coformer screening makes traditional, intuition-driven approaches slow and incomplete.

This white paper details the development and validation of an AI modeling tool for cocrystal screening, drawing on perspectives from cheminformatics data scientist Aaron Johnson and preformulation scientist Kelly Shunje. The piece covers the team's methodology in building a training set of over 20,000 data points, the shift from anomaly detection to supervised machine learning, and an ensemble modeling approach that correctly predicted experimental outcomes in approximately 8 of 10 held-out cases. It also addresses a consideration that often goes unexamined in AI-assisted drug development: what happens to proprietary molecular structures during the modeling process.

For scientists and development teams thinking about when and how to integrate computational tools into solid form strategy, the full white paper is worth examining.