Fast energy minimization of the CCDC drug-subset structures by molecule-in-cluster computations allows independent structure validation and model completion

Abstract

Optimizing structures with computations on clusters of molecules permits generation of structure-specific restraints for refinement. Equally importantly, retrospective structure validation and addition of hydrogen atoms consistent with quantum chemistry is possible for experimental structures or the solvent molecules in them, should they be missing in earlier CIF depositions. Revisiting the drug subset structures of the CCDC demonstrates that structure validation through ab initio cluster computations is a tremendous validation tool. The time required for optimization can be similar to the time required to carry out least-squares refinement for small-molecule structures, and becomes feasible for large structures. Several questions arise: is it valid to augment experimental structures with structure-specific restraints, ideally through accompanying refinement with computation? Do energy minimized structures (using the experimental determinations as a starting point) still constitute an experimental result? When re-refinement is impossible in retrospect, like for most of the drug-subset molecules, then additional value lies in completion and validation of existing structures so that they are chemically and crystallographically correct, and contain missing water or solvent hydrogen atoms. Our results suggest that retroactive validation and addition of hydrogen atoms becomes possible for the entire Cambridge Structural Database. Generation of database entries of optimized alongside existing structures will provide the flexibility needed to make full use of the information gained by computation.