A structural keystone for drug design

Abstract

3D-structures of proteins and potential ligands are the cornerstones of rational drug design. The first brick to build upon is selecting a protein target and finding out whether biologically active compounds are known. Both tasks require more information than the structures themselves provide. For this purpose we have built a web resource bridging protein and ligand databases. It consists of three parts: i) A data warehouse on annotation of protein structures that integrates many well-known databases such as Swiss-Prot, SCOP, ENZYME and others. ii) A conformational library of structures of approved drugs. iii) A conformational library of ligands from the PDB, linking the realms of proteins and small molecules.

The data collection contains structures of 30,000 proteins, 5,000 different ligands from 70,000 ligand-protein complexes, and 2,500 known drugs. Sets of protein structures can be refined by criteria like protein fold, family, metabolic pathway, resolution and textual annotation. The structures of organic compounds (drugs and ligands) can be searched considering chemical formula, trivial and trade names as well as medical classification codes for drugs (ATC). Retrieving structures by 2D-similarity has been implemented for all small molecules using Tanimoto coefficients. For the drug structures, 110,000 structural conformers have been calculated to account for structural flexibility. Two substances can be compared online by 3D-superimposition, where the pair of conformers that fits best is detected. Together, these web-accessible resources can be used to identify promising drug candidates. They have been used in-house to find alternatives to substances with a known binding activity but adverse side effects.