Mapping of Protein Binding Sites using clustering algorithms - Development of a pharmacophore based drug discovery tool
Jessica L. Braun, Darren Fayne
Abstract
Discovering new hit small molecules binding to a specific protein binding site can be a difficult task. In support of existing procedures, a proof of concept methodology has been developed to process fragment flooded X-ray protein structures using the K-means clustering algorithm in order to derive pharmacophore models of the binding site. The novel method includes the implementation of several K-means initialisation methods in serial and parallel versions. Furthermore, required parameter optimisations for two initialisation methods was achieved, which was necessary to determine their validity and performance. A graph theory algorithm was adapted to compare the clustering-derived pharmacophores with X-ray ligand structure-derived pharmacophores to confirm that they mapped to each other. Initial proof of concept method validation was demonstrated using the Androgen Receptor (AR).