Abstract
A computerized retrieval system is described which allows the interactive identification of pharmacophoric patterns in 3D chemical structures. A query pharmacophore in the system is matched against a file of structures using a three-stage retrieval mechanism. The initial search involves a rapid bit-string matching operation using screens denoting the presence or absence of interatomic distances that lie within certain predefined ranges. These ranges are selected by an analysis of the frequencies of occurrence of interatomic distances in the data file that is being searched. The second stage involves the generation of all interatomic distances in potentially matching structures in the file to ascertain whether all of the query distances are present. If this is the case, a final search is carried out to determine whether these distances are in the correct geometric orientation to each other; this search involves a novel matching algorithm based upon the set reduction techniques used for substructure searching in files of 2D chemical structures. The evaluation of the system has involved the searching of several pharmacophoric patterns against a 12 728-compound subset of the Cambridge Crystallographic Data Bank. Measurements of the numbers of structures matching at each of the three levels of search show that the system provides a highly efficient means of identifying structures containing a query pharmacophore.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call