Atom pair 2D-fingerprints perceive 3D-molecular shape and pharmacophores for very fast virtual screening of ZINC and GDB-17

J Chem Inf Model. 2014 Jul 28;54(7):1892-907. doi: 10.1021/ci500232g. Epub 2014 Jul 2.


Three-dimensional (3D) molecular shape and pharmacophores are important determinants of the biological activity of organic molecules; however, a precise computation of 3D-shape is generally too slow for virtual screening of very large databases. A reinvestigation of the concept of atom pairs initially reported by Carhart et al. and extended by Schneider et al. showed that a simple atom pair fingerprint (APfp) counting atom pairs at increasing topological distances in 2D-structures without atom property assignment correlates with various representations of molecular shape extracted from the 3D-structures. A related 55-dimensional atom pair fingerprint extended with atom properties (Xfp) provided an efficient pharmacophore fingerprint with good performance for ligand-based virtual screening such as the recovery of active compounds from decoys in DUD, and overlap with the ROCS 3D-pharmacophore scoring function. The APfp and Xfp data were organized for web-based extremely fast nearest-neighbor searching in ZINC (13.5 M compounds) and GDB-17 (50 M random subset) freely accessible at .

Publication types

  • Research Support, Non-U.S. Gov't

MeSH terms

  • Databases, Pharmaceutical*
  • Drug Evaluation, Preclinical / methods*
  • Informatics / methods*
  • Ligands
  • Molecular Conformation*
  • User-Computer Interface


  • Ligands