Clustering of binding solutions for computer aided drug design

Abstract

Amyotrophic lateral sclerosis (ALS) is a disease presenting with severe, progressive neurodegeneration. Binding of the retinoic acid receptors (RAR) and retinoic X receptors (RXR) with retinoids may promote neuroprotection and neurite outgrowth. With potential applications in ALS in mind, a method for screening potential synthetic retinoid drug candidates was developed in which binding complexes for the candidates are predicted in a statistical manner.
In this thesis, the statistical method of binding prediction is called clustering. Drug candidates were docked in silico and a python script was developed to group the multiple geometric solutions from docking into clusters of similar looking solutions.
The python script could reproduce the binding poses of synthetic retinoids, for which binding structures have already been resolved, with good accuracy and consistency. This new tool can be used to predict probable geometries of ligands in binding domains that have not yet been solved crystolographically, thus establishing a rapid understanding of how interactions may occur in such complexes without even entering a laboratory.