The Overprediction Problem: Aligning Computational Crystal Structure Prediction with Experimental Reality for Two Sulfa Drugs
Year: 2026
Type: article
Abstract: Computational crystal structure prediction (CSP) methods usually overpredict the number of plausible polymorphs. In this study, we used sulfadiazine (SDZ) and sulfamerazine (SMZ), which are respectively monomorphic and pentamorphic (by the end of this study), to investigate the extent of CSP overprediction and its use in designing experiments that identify additional polymorphs. Lattice energy CSP studies (CSP_0) generate a few hundred plausible crystal structures (low-energy minima) for these molecules; however, if these crystal structures are simulated at ambient conditions using a molecular dynamics workflow, then many of the structures can be clustered as equivalent, reducing the number of distinct structures by about 60%. The relative energies are refined using a variety of periodic DFT+D and phonon calculations on the competitive structures. Comparison of the known and CSP-generated thermodynamically competitive structures suggests that forming solid solutions or subliming onto template crystals could enable the discovery of new forms. Solid solutions with the predicted structure of SMZ-IV were found, with SDZ content ranging from 30% to 70%. The combined computational and experimental study suggests that SDZ is monomorphic and that SMZ form II is the low-temperature stable form, enantiotropically related to the polymorphs I, III, and V of SMZ due to the unusual thermal stabilization of a layer structure that occurs in the most readily crystallized form I. The contrasting polymorphisms of SMZ and SDZ illustrate the reasons for overprediction and the requirements for polymorph prediction by CSP.
DOI: 10.1021/jacs.6c14137
Institutions: University College London, Pfizer (United States)
Language: English
Subfield: Materials Chemistry
Field: Materials Science
Domain: Physical Sciences
Open Access status: hybrid
Estimated APC paid: $4,500
OpenAlex ID: W7236154412
