Raymond E. Carhart

Managing the Combinatorial Explosion

Computer software designed to deal with the large amounts of data generated by chemical and biological research programs is described. The software includes methods for representation of the structural components of combinatorial libraries, enumeration of structures and structure-based searches, and comparisons within combinatorial libraries.

A Model-Based Approach to the Teletype Printing of Chemical Structures

A Fortran program for drawing chemical structures on the teletype starting from a connection table is described. The program is guided in its task by an internally constructed model of the molecule and is thus freed from the limitations of template-based systems. An outline of the program logic is presented along with an example. Several samples of the program output are included to indicate its general performance level, and the availability of the program is discussed.

Erroneous Claims Concerning the Perception of Topological Symmetry

A recent issue of this journal carried two dealing directly or indirectly with “inexpensive” (in the sense of low computational effort) methods of perceiving molecular ~ y m m e t r y . ~ More precisely, each article gives a set of rules for scoring2 or comparing) atoms in a molecule with the claim that if the scores are equal, or if the comparison shows no difference, then the atoms are symmetrically equivalent (Le., can be interchanged by some symmetry operation on the molecule). Though both methods are doubtless very good approximations in the sense that they almost always yield the classes of symmetrically equivalent atoms for typical chemical molecules, it is demonstrated below that neither is fully correct. There exist chemical graphs for which the number of distinct symmetry types of atoms is greater than that computed by the suggested algorithms. If these approximate methods are taken as accurate and used, as indicated by Shelley and Munk,2 “. . . in applications of I3C N M R spectroscopy to chemical problems and in the canonical representation and elaboration of molecular structures. . .”, then it must be accepted that in some instances the wrong number of 13C N M R peaks may be predicted, nonunique “canonical” numberings may be created, and correct structures may be overlooked during elaboration. In developing programs of this kind, it is important for one to support each new algorithm for manipulating chemical structures with a solid foundation of graph-theoretic understanding, lest unexpected and perhaps unnoticeable errors occur in the output. Verifying that an algorithm is never known to fail when its output is compared with selected manually generated cases, or with the output of other programs based on totally different principles, does not constitute proof that the algorithm is valid beyond the specific cases which are tested. Such a “proof’ leaves open the important question of the scope and limitations of the algorithm: How far beyond those test cases would one have to look to find one which fails? For this reason, my co-workers and I have, wherever possible, provided rigorous graph-theoretic proofs supporting the methods used in the DENDRAL program^.^,^ It is crucial that readers of this journal understand the approximate nature of the cited symmetry-perception algorithms before applying the methods to their own problems.

Structural isomerism of mono- and sesquiterpenoid skeletons☆☆☆

Abstract A generator of chemical structures (CONGEN) has been utilized to investigate two aspects of the structural isomerism of mono- and sesquiterpenoid skeletons: (1) the scope of possible isomers under various structural constraints; and (2) the scope of possible isomers based on a mechanistic model which allows interactive exploration of reactions of formation and interconversion. The possibilities, even under severe constraints, are many more than the structural types commonly encountered in natural. These results indicate the potential danger of structural assignment based in part on biogenic grounds.

Applications of artificial intelligence for chemical inference—XX: “Intelligent” use of constraints in computer-assisted structure elucidation☆☆☆

Abstract The CONGEN program for computer-assisted structure elucidation constructs possible solutions to a structure elucidation problem from atoms and other structural units supplied by the chemist. Most such problems have many constraints on the plausibility of certain types of structures. These constraints are also supplied to the program. But constraints expressed in chemical terms cannot always be easily translated into the more rigid graph-theoretical domain of the structure generator. The current methods of implementation of constraints in CONGEN are described in terms of the strategies we have adopted to emulate a chemists reasoning about molecular structures.

Applications of artificial intelligence for chemical inference. Part XI. Analysis of carbon-13 nuclear magnetic resonance data for structure elucidation of acyclic amines

This paper describes a computer program entitled AMINE, which uses a set of predictive rules to deduce the structures of acyclic amines from their empirical formulae and 13C n.m.r. spectra. The results of testing the program on 102 amines indicate that AMINE is quite accurate and selective, even for large amines with many millions of structural isomers, and demonstrate that the computerized analysis of 13C n.m.r. data can be a powerful analytical tool. The logical structure of the program is outlined here, including a section on the general problem of spectrum matching. Generalizations of the methods used by AMINE are suggested.

Information Integration: Distributed Chemical Information Management Systems

The new generation of distributed computing systems offers exciting new ways to bring together chemical information, including chemical structures and reactions, data, text and graphics. The connection of high performance workstations to networks of geographically distributed computers rather than to single hosts provides the connectivity to support information integration. The next challenge is to construct software systems that actually deliver the potential functionality. We will discuss software architectures designed to achieve integration at both the workstation and host. End-user control over how information is accessed and presented, independent of its geographical location, will be shown to be an essential part of such systems. Emerging standards for transport of information among diverse applications will be discussed as the ‘glue’ that makes distribution the enabling technology for integration.