Atom-Atom Mapping
Introduction
Atom-Atom mapping in a chemical DBMS is an approach whereby the chemical DBMS is informed that an atom in the reactant corresponds to an atom in the product. Most often the atoms mapped are reaction centres. The DBMS query is performed on a reaction database with SQL database operator arguments providing one means of confering the atom-atom mapping between the reactant and reaction product. The following simple SNAr reaction of a 2-bromopyridine with a metal amide is a very simple example of atom-atom reaction centre mapping - the cyclic N in the reactant remains the cyclic N in the product [ie the NH2 is substituted for the Br in the product (reaction mechanism not included here)].
It has been experimentally proven that this atom-atom mapping is correct. A heterocyclic chemistry specialist might intuitively expect this mapping too. Experimental results from isotopic labelling studies confirming this mapping, if available for the exact or similar reaction scheme, are however infrequently included in and used by COTS DBMS chemical systems.
Atom-Atom mapping in a chemical DBMS is usually a nonsense
A Google search of the term atom-atom mapping produces many hits. The functionality touted most often however does not exist in a chemically correct and meaningful manner.
The atom-atom mapping in the metal amide substitution reaction shown above happens to be correct. However in most chemical DBMS systems, the logic is nothing more than pattern matching and often a chemical nonsense.
The primary problem with atom-atom mapping is the assumption that skeletal atom connectivity is preserved. The chemist or chemoinformatics practioner may not know this in advance, the database containing this reaction detail may not be populated with chemical rearrangement information, and/or the reaction mechanism of the chemical transformation may not be known. Without isotopic labelling studies of the specific reaction in question, the atom-atom mapping is often just pattern matching lacking scientific foundation.
A reaction very similar to the metal amide substitution of a N-heteroaromatic compound shown above is discussed briefly.
In this apparently trivial substitution reaction, the mapping between the N of the reactant and the cyclic N of the reaction product is similarly represented by the coloured line. In this specific reaction, a poorly implemented chemical DBMS would allow the chemoinformatic practioner to believe that that there exists a chemically significant mapping between the N reaction centres. None exists - the chemical DBMS implementation that returns this reaction as part of the result set is is most likely to be pattern matching. In this specific reaction, a molecular rearrangement has taken place and the cyclic N in the reactant has been shown to be the exocyclic N in the reaction product! The reader is referred to the ANRORC (Addition of Nucleophile, Ring Opening and Ring Closure mechanism); see Van Allan, J. A., Reynolds, G. A., Maier, D. P., and Chang, S. C., J. Heterocycl. Chem., 1972, 9, 1229. This reaction mechanism is also discussed in Heterocyclic Chemistry, third edition, Joule, J. A., Mills, K., and Smith, G. F., ISBN 0-7847-4069-4, page 125).
Summary
Atom-atom reaction mapping is often a useful mining process in chemical DBMS implementations. The mapping implemented within the DBMS logic is however most likely to be nothing more than pattern matching. Chemists and chemoinformatic practioners will have to ensure that the database result set returned using such technology is chemically valid, and that the result set excluded through the use of this technology has not been excluded in the basis of a misunderstanding or ignorance of simple chemical principles.