Adding value to crystallographically-derived knowledge bases.

A protocol for the partially automated computational investigation of crystal structure geometries of transition-metal complexes with unusual/outlier structural features has been developed for application in an e-science context. This protocol not only is envisaged as a part of knowledge base software packages such as Mogul but can also be used to further analyze the results of database searches. The issues arising from automating the initial input generation and DFT optimization of complexes have been examined and a procedure for extracting additional knowledge "value" from the computational results is described. Potential problems/weaknesses arising from the choice of computational approach and from errors in the crystal structure refinement are discussed. A range of likely outcomes of applying this protocol to database mining results is illustrated, with representative examples identified for tetracoordinate transition-metal complexes and ligand fragments (terminal chloride, monodentate phosphorus(III), and primary amine ligands) with unusual metal-ligand bond lengths.