Co-crystal Prediction by Artificial Neural Networks*.

Area of Science:

• Materials Science
• Computational Chemistry
• Crystallography

Background:

• Co-crystals offer tunable physicochemical properties crucial for drug development and materials science.
• Identifying suitable coformer pairs for co-crystal formation remains a significant challenge, limiting solid-state landscape exploration.

Purpose of the Study:

• To develop and apply a data-driven prediction method for identifying potential co-crystal forming pairs.
• To overcome the limitations in efficiently exploring the co-crystal solid-state landscape.

Main Methods:

• Utilized two types of artificial neural network (ANN) models trained on co-crystal data from the Cambridge Structural Database.
• Developed a method that predicts the likelihood of co-crystal formation for given pairs of coformers.

Main Results:

• The combined output of multiple ANN models demonstrated excellent performance on training and validation datasets.
• Achieved an estimated accuracy of 80% for predicting co-crystallization for novel molecular pairs lacking prior data.

Conclusions:

• The data-driven ANN approach effectively predicts co-crystal formation, addressing a key challenge in materials science.
• This method facilitates efficient exploration of the solid-state landscape, accelerating the discovery of new co-crystals.