COMPUTER STORAGE AND RETRIEVAL OF GENERIC CHEMICAL STRUCTURES IN PATENTS .16. THE REFINED SEARCH - AN ALGORITHM FOR MATCHING COMPONENTS OF GENERIC CHEMICAL STRUCTURES AT THE ATOM-BOND LEVEL

摘要

A description is given of an atom-level search strategy for matching components of generic chemical structures, called the refined search. The components used are those represented by nodes of a reduced graph. These nodes describe aggregates of atoms of the original chemical graph which are similar in structure or chemistry. The reduced graph screening-search results in a List of pairs of matched nodes, one query structure node and one file structure node in each pair, which is the basis for the refined search. The atoms and bonds of a query structure node are compared with those of a file structure node. Due to the nature of generics, parts of the structure are expressed either in terms of real atoms and bonds or in terms of homologous series identifiers, such as alkyl. The fundamental problem concerning an atom-level search is the translation between the two types of representation for these different expressions. An adaptation of the Ullmann algorithm for substructure isomorphism has been devised to solve this problem; this is described together with several example matches. [References: 23]