A robust crystal structure prediction method to support small molecule drug development with large scale validation...

Area of Science:

• Solid state chemistry
• Crystallography
• Materials science

Background:

• Crystal polymorphism significantly impacts pharmaceuticals, agrisciences, and other industries.
• Late-appearing, more stable polymorphs pose challenges in drug development.
• Experimental polymorph screening is costly, time-consuming, and may miss low-energy forms.

Purpose of the Study:

• To report a highly accurate and efficient crystal structure prediction (CSP) method.
• To validate the CSP method on a large, diverse dataset of known polymorphic forms.
• To demonstrate the method's utility in identifying potential new polymorphs and aiding formulation design.

Main Methods:

• A novel systematic crystal packing search algorithm.
• Machine learning force fields for hierarchical crystal energy ranking.
• Validation on 66 molecules with 137 experimentally known polymorphic forms.

Main Results:

• The CSP method successfully reproduced all experimentally known polymorphs.
• The method predicted new, low-energy polymorphs not yet discovered experimentally.
• Successful prediction in a blinded study and the seventh CSP blind test was achieved.

Conclusions:

• The developed CSP method offers state-of-the-art accuracy and efficiency.
• This tool can identify potential risks from undiscovered polymorphs.
• The method accelerates clinical formulation design and mitigates downstream processing risks.