Will it Crystallise

What is it about?

For a crystallographer interested in molecular structure, the biggest barrier to studying materials is obtaining them in a crystalline form. Machine learning approaches have previously been applied to predict many properties of molecules – in this paper we show that there is sufficient information in a conventional 2D chemical diagram (atom types and bonds) to accurately predict whether a material crystallises easily.

Featured Image

Why is it important?

A unique aspect of the approach described is the corralling of large data sets (CSD and ZINC) into forms which can be used to answer a specific problem. The predictions made by this model can be used to inform recrystallisation experiments in the lab, or to guide small synthetic modifications to molecules in order to alter crystallinity of a material.