IVGA3D: De Novo Ligand Design Using a Variable Sized Tree Representation

Sanghamitra      Bandyopadhyay; Soumi      Sengupta

doi:10.2174/0929866511009011495

Abstract

Rational drug design is an important step towards effective patient care enabling lead molecule discovery in a relatively faster and inexpensive way. This article describes a variable string length genetic algorithm with domain specific operators for de novo ligand design. The algorithm first mines the active site of the given protein receptor whose geometry and chemical composition guides the ligand building. Active site mined by the algorithm is compared with two more established active sites detecting schemes to evaluate its efficiency. Various combinations from a suite of forty one fragments are mined to design the ligands. Bond stretching, angle bending, torsional terms, van der Waals and electrostatic interaction energy with distance dependent dielectric constant contribute are used to compute the internal energy of the ligand and the interaction energy of the ligand receptor complex. Forty one fragments are used to design the ligands. Experimental results are provided for HIV-1 Protease, HIV-1 Nef and Thrombin demonstrating the superiority of the proposed scheme vis-a-vis three other approaches. Comparison with known inhibitors also demonstrates the effectiveness of the proposed approach.

Keywords: Drug design, active site, de novo ligand design, genetic algorithms, Rational drug design, HIV-1 Protease, HIV-1 Nef, Thrombin, X-ray crystallography, NMR spectroscopy, RNA polymerase, major histocompatibility complex, HIV infection, octapeptide drug, SARS, calmodulin, chrysin, QSAR, LEA3D, variable length genetic algorithm, Protein Data Bank, building block hypothesis, IVGA3D, Non-bonding interactions, Decoding a Chromosome, chromosome, polar hydroxyl group, hydrophobic amino acids, methionine, Elitism, CASTP, Q-SiteFinder, ConQuest, Food and Drug Administration, (FDA), (RMSD), Cambridge Structural Database