Prediction and Preparation of Coamorphous Phases of a Bislactam

The effectiveness of a partial least squares-discriminant analysis coamorphous prediction model was tested using coamorphous screening data for a promising coamorphous former, the dimer of N-vinyl(caprolactam) (bisVCap) with a range of active pharmaceutical ingredients. The prediction model predicted 71% of the systems correctly. An experimental coamorphous screen was performed with this coformer with 13 different active pharmaceutical ingredients, and the results were compared to the predictions from the model. A total of 85% of the systems were correctly predicted. Stability assessments of three coamorphous systems showed that the prediction model score did not strongly correlate with the stability of the coamorphous material. The model performed well with small-molecule co-formers, such as bisVCap, despite the difference in structure and properties compared to the amino-acid-based model training set.