One of the golden rules of cheminformatics is that tools should preserve chemical information exactly as provided by the user. Despite this, for some applications it may be useful to derive from the chemical structure a representation that discards or normalizes some of that information. We can refer to these as a molecular hash (or ‘molhash’) by analogy with hash functions from computer science. The use of molecular hashes presents an efficient route to solving many problems in cheminformatics which otherwise require exhaustive enumeration. The most well-known molecular hash is the molecular formula, e.g. C9H8O4 for aspirin. This is a representation of the structure that discards connectivity information and stereochemistry, but is sufficient to calculate the molecular weight. If particular rules are followed for the generation of the formula (e.g. by ordering the elements in Hill order) then the molecular formula is a convenient way to sort chemical structures so that they can easily be looked up; the Merck index has a formula index, for example. As the molecular formula ignores bond information, it is invariant to differences in bond representation common among inorganic complexes. Furthermore, by collating molecules based on their molecular formula, it is possible to identify isomers, which may be constitutional isomers or stereoisomers. The example of the molecular formula illustrates that even the simplest of molecular hashes can be useful in identifying molecules that are structurally related in some way. This presentation will illustrate the application of molecular hashes to finding stereoisomers, alternate resonance forms, matched pairs, tautomers, regioisomers, redox pairs, mesomers, molecules that differ by a single-atom substitution, molecules with the same graph layout, and related peptides. We will discuss the concept of a hierarchical relationship between hashes, which leads to the concept of parent forms, and how calculation of slow hashes can be avoided by first checking parent forms.

Making a hash of it: Advantage of selectively leaving out structural information

2019 Fall National Meeting

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.