Three-dimensional quantitative structure-activity relationships from molecular similarity matrices and genetic neural networks. 2. Applications

J Med Chem. 1997 Dec 19;40(26):4360-71. doi: 10.1021/jm970488n.


Validation of a method that uses a genetic neural network with electrostatic and steric similarity matrices (SM/GNN) to obtain quantitative structure-activity relationships (QSARs) is performed with eight data sets. Biological and physicochemical properties from a broad range of chemical classes are correlated and predicted using this technique. Quantitatively the results compare favorably with the benchmarks obtained by a number of well-established QSAR methods; qualitatively the models are consistent with the published descriptions on the relative contribution of steric and electrostatic factors. The results demonstrate the general utility of this method in deriving QSARs. The implication of the importance of molecular alignment and possible methodological improvements are discussed.

Publication types

  • Research Support, U.S. Gov't, Non-P.H.S.

MeSH terms

  • Cholinesterase Inhibitors / chemistry
  • Dopamine beta-Hydroxylase / antagonists & inhibitors
  • Drug Design
  • Enzyme Inhibitors / chemistry
  • Enzyme Inhibitors / metabolism
  • GABA-A Receptor Agonists
  • GABA-A Receptor Antagonists
  • Molecular Conformation
  • Molecular Structure
  • Neural Networks, Computer*
  • Phosphorylases / antagonists & inhibitors
  • Receptors, Aryl Hydrocarbon / metabolism
  • Static Electricity
  • Structure-Activity Relationship*


  • Cholinesterase Inhibitors
  • Enzyme Inhibitors
  • GABA-A Receptor Agonists
  • GABA-A Receptor Antagonists
  • Receptors, Aryl Hydrocarbon
  • Dopamine beta-Hydroxylase
  • Phosphorylases