In silico classification of adenosine receptor antagonists using Laplacian-modified naïve Bayesian, support vector machine, and recursive partitioning

J Mol Graph Model. 2010 Jun;28(8):883-90. doi: 10.1016/j.jmgm.2010.03.008. Epub 2010 Mar 27.

Abstract

Adenosine receptors (ARs) belong to the G-protein-coupled receptor (GPCR) superfamily and consist of four subtypes referred to as A(1), A(2A), A(2B), and A(3). It is important to develop potent and selective modulators of ARs for therapeutic applications. In order to develop reliable in silico models that can effectively classify antagonists of each AR, we carried out three machine learning methods: Laplacian-modified naïve Bayesian, recursive partitioning, and support vector machine. The results for each classification model showed values high in accuracy, sensitivity, specificity, area under the receiver operating characteristic curve and Matthews correlation coefficient. By highlighting representative antagonists, the models demonstrated their power and usefulness, and these models could be utilized to predict potential AR antagonists in drug discovery.

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Algorithms
  • Artificial Intelligence*
  • Bayes Theorem
  • Computer Simulation
  • Drug Design
  • Models, Chemical*
  • Molecular Structure
  • Purinergic P1 Receptor Antagonists*
  • Structure-Activity Relationship

Substances

  • Purinergic P1 Receptor Antagonists