A HYPERSTRUCTURE MODEL FOR CHEMICAL-STRUCTURE HANDLING - GENERATION AND ATOM-BY-ATOM SEARCHING OF HYPERSTRUCTURES

This paper discusses the representation of sets of chemical structures by hyperstructures, pseudomolecules in which areas of structural commonality are stored nonredundantly. Two methods are described for the construction of hyperstructures, and it is shown that one of these methods, the atom-assignment method, is far less demanding of computational resources than the other. Experiments with a file of 10K structures demonstrate that the hyperstructure for a dataset requires less storage than the original set of structures; however, the difference is not large. The experiments also demonstrate that hyperstructures have at least some potential for increasing the speed of atom-by-atom searching, as compared with conventional chemical database systems.