3D Molecular Similarity: Method and Algorithms

Oleg URSUa*, Mircea V. DIUDEAa and Shin-ichi NAKAYAMAb

aFaculty of Chemistry and Chemical Engineering, 11, Babes-Bolyai University
Arany Janos Str., 400028 Cluj-Napoca, ROMANIA
bUniversity of Tsukuba, Research Center for Knowledge Communities
1-2 Kasuga, Tsukuba, Ibaraki, 305-8550, JAPAN

(Received: May 3, 2006; Accepted for publication: October 24, 2005; Published on Web: March 2, 2006)

This study presents a method and algorithms for calculation of 3D similarity between pairs of chemical structures represented as 3D molecular graphs. Similarity searching in chemical databases is widely used for virtual screening, lead discovery and optimization, and most recently protein amino-acid sequences studies to discover and determine the functionality of a new isolated protein. This method has obvious advantages over other known methods due to the following: (i) the superposition method does not depend on the preliminary alignments of the chemical structures; (ii) entire conformational space is searched without generation of each conformer; (iii) excellent discrimination between geometrical isomers. Although it is a computationally demanding method, recent implementation of maximum clique algorithm and bound smoothing algorithm made possible the optimization of this method and application to similarity searching in chemical databases of non trivial size.

Keywords: Molecular similarity, Maximum common subgraph, Distance geometry, Similarity coefficient

Text in English

PDF file(408kB)