Pharmacoinformatics Guided Virtual Screening Pipeline to Mimic hERG Inhibition Profiles of New Chemical Entities

Abstract

One of the major causes of sudden cardiac death is Torsades de pointes (TdP), a polymorphic tachycardia with characteristic “twisting of the points” on the electrocardiogram. Torsades de pointes is frequently associated with acquired long QT syndrome. Several experimental studies have provided evidence of QT prolongation due to drug induced blockade of human ether-a-go-go (hERG) potassium channel. The hERG channel represents important component of cardiac action potential (CAP) and is responsible for the rapid component of delayed rectifier current (IKr). Drug induced blockade of the hERG channel has been associated with the highly promiscuous properties of its binding site. Therefore, it can trap a range of structurally and functionally diverse compounds in its inner cavity. In the last 15 years, several drugs have been withdrawn from the market due to cardiac toxicity. This necessitates the development of reliable in-silico models to predict hERG blockade in early phases of drug development. Therefore, this dissertation aims to develop combined ligand and structure guided hERG liability predictive protocol to probe the 3D structural features of chemical entities involved in the blockade of hERG. The study also aims to correctly rank the NCEs in different potency (nM-μM) levels against hERG. Towards these goals, several 3D QSAR Grid INdependent Descriptor (GRIND) models have been built using different 3D conformational sets of structurally diverse data set of hERG inhibitors. Ouerall, predictive ability and robustness of models helped in correctly rank the potency order (lower μM to higher nM range) against hERG. A small deviation of about ~0.4Å was observed between important hotspots of molecular interaction fields (MIFs) between solvated and non-solvated hERG models. These small changes in conformations do not affect the performance and predictive power of the model to any significant extent. Therefore, neglecting water molecule does not have a significant implication on the computational models for the prediction of hERG inhibition potential but it will help in reducing computational cost and time. The lipophilic and ligand efficiency profiling was performed to identify ligands with best potency/lipophilicity and optimal fit within the binding site. The most Abstract xiii probable binding conformations of selected ligands were used for the ligand-protein interaction guided pharmacophore models. Finally selected Pharmacophore model highlighted the presence of two aromatic one hydrophobic and one hydrogen bond acceptor group at particular mutual distance in most potent hERG inhibitors which complement the respective hydrophobic and hydrogen bond donor contours at the virtual receptor site produced by the final GRIND model. Selected Pharmacophore was used for virtual high-throughput screening of different publically available databases. Finally, selected hits were experimentally validated using whole-cell patch-clamp technique. The experimental validation revealed a difference of less than ±1.6 log unit between experimentally determined and predicted hERG inhibition potential (IC50) of the selected hits. Overall, this work could help in identifying hERG inhibitors during early phase of drug development to prevent cardio toxicity and pave the way towards safer drug development.