On of molecular similarity based on a single molecular descriptor

We consider the characterization of molecular similarity through a single molecular descriptor, instead of the customary use of sets of structural invariants to characterize individual molecules. Moreover, we require that the ‘similarity’ descriptor be conceptually and computationally simple so that it is suitable for screening huge combinatorial libraries in search for target compounds. We have outlined one such general approach for construction of ‘similarity’ descriptors, which is illustrated on the set of 35 nonane constitutional isomers.