Substructures Searching Generic

The CDD staff concluded that MACCS could not rephcate the registration process of CSIS. It also did not provide the sophisticated generic substructure searching capabilities which are used in a least 45% of the searches executed by the CDD staff. The absence of an interface with the department s ancillary databases was another serious deficiency of MACCS. 

Markush searches are generic substructure searches. It should be clear that both structure queries and structure files may contain specific and/or generic Structures. In a Markush search generic chemical structures are searched against a database which contains specific or generic structures. The first type of database is a registry file, e.g., the CAS REGISTRY, and the second type is a Markush file, e.g., MaRPAT (see Structure Representation). 

Chemical data base systems, and especially the DARC Structure Management System (DARC-SMS), have from the beginning offered sophisticated capabilities for chemical structure registration and structure, or substructure, or generic substructure search. 

Fig. 7. Nontrivial overlap between two simple com- Fig. 8. Substructure search of generic structures, binatorial libraries.

Another searching capability is required by the need to find a partial structure or substructure in a database that includes generic structures. The ability to do a substructure search is standard in modem chemical structure information systems. The problem is more complicated when generic structures must be searched, yet even more necessary since the generic might represent millions of specific structures. The query in Fig. 8 is compared with the generic structure to find the set of specific structures. 

Substructure searching methods, then as now, retain pride of place in interest. Here, the advances are especially strongly evident many of the papers this week report on the operations of in-house and public structure and data retrieval systems, now very widespread in industry. Fourteen years ago, the ICI and BASF systems were two of a few operational search systems in industry, with Richard Feldmann s forerunner of the Chemical Information System(CIS) at the National Institutes of Health. The ICI CROSSBOW system, though, depended for some of its functions on the use of notations at the user interface, which restricted its use somewhat, as chemists wereunwilling to learn to encode structures for input. The performance of the BASF system was remarkable for the time, yet had already been in operation for over five years, and even included some generic structure storage and search capabilities. 

Gillet, V.J. Welford, S.M. Lynch, M.F. Willett, P. Barnard, J.M. Downs, G.M. Manson, G. Thompson, J. Computer Storage and Retrieval of Generic Chemical Structures in Patents. 7. Parallel Simulation of a Relaxation Algorithm for Chemical Substructure Search. J. Chem. Inf. Comput. Sci. 1986, 26, 118-26. 

Generic structxires pose problems which far outstretch the capabilities of topologically-based systems for substructure search of specific chemical substances. As the example of Figure 1 shows, they are framed in terms of characteristics which include the following ... 

Certain of these features, notably the first type, together with don t care bond types, have already appeared in operational substructure search systems, as amplifications to the query language, to simplify the users task. They represent a small subset of the features which occur in generics in patents. 

It is conceivable that extensions to existing substructure search systems, and a good deal more computing power, would be sufficient to deal with searches of generics which include the first four of these features. Extensions not involving new knowledge to deal with the other types can be ruled out. 

In chemical structure terms, he may wish to search for an exact structure of a reactant or product (although this is probably more useful with an in-house than with a literature database) or for a substructure in a reactant or product. However, what he would most like to do is generic reaction search. 

The underlying composite generic and specific structure (see Section 3.3.2.1(b) above) allows for fuzzy structure search. The ability to tag portions of a query structure is limited to a generic group basis. On the basis of such portions in fuzzy structure search, one can range from the exactness of substructure search to the fuzziness of a similarity search (the default). 

---