Development of an in silico predictive protocol for hERG liability

COMP 192

Hongwu Wang1, Vincent Madison1, Xue-Song Zhang2, and Steve Sorota2. (1) Department of Structural Chemistry, Schering-Plough Research Institute, 2015 Galloping Hill Road, Kenilworth, NJ 07033, (2) Department of CNS-CV Research, Schering-Plough Research Institute, 2015 Galloping Hill Road, Kenilworth, NJ 07033
Inhibition of the human Ether-a-go-go Related Gene (hERG) ion channel is associated with prolongation of the QT interval of the surface electrocardiogram, reflecting delayed repolarization of the heart. Compounds exhibiting this action may predispose patients to ventricular arrhythmias and sudden death. Because many structurally diverse compounds have been found to inhibit the hERG channel, limiting the hERG liability has become a major hurdle for many drug discovery programs. We developed a computational protocol for the prediction of hERG inhibition using screening data from a rubidium efflux assay on 2695 Schering compounds from 20 drug discovery projects. The protocol consists of two independent models: a categorical model developed using a Bayesian method and a similarity search method that takes advantage of the growing experimental database. Each model by itself can correctly predict the hERG activity for 70%-80% of the potent and weak inhibitors in a test set. The combination of these two models can make very reliable predictions for a large portion of the compounds. Issues regarding the quality of the screening data and structural features critical for hERG inhibition will also be discussed. This protocol can be employed as a virtual screening tool in drug discovery to reduce hERG liability.