The human ether-a-go-go related gene (hERG) blockade associated cardiovascular toxicity is a common reason for drug failure in the preclinical safety studies and its assessment is mandatory for the regulatory approval of drugs. Early prediction of hERG liability of compounds in drug discovery program is critical. A total of 961 compounds with the measured hERG activity collected from ChEMBL database were used to build binary kernel classification models, discriminating hERG actives from inactives. Models with a different combination of descriptors comprised of molecular fingerprints, chemical properties and APF based continuous pharmacophoric 3D descriptors were generated. Preliminary models, with the accuracy higher than 0.9 and 0.8 for training and test sets, respectively, were rigorously validated and applied to predict the hERG liability of compounds from challenging validation sets. The greater than 70% prediction accuracy of external compounds highlights the effectiveness of these models in screening of novel chemotypes for hERG liabilities.

Development of ligand-based models for predicting hERG toxicity of chemicals

243rd National Meeting (2012)

The official archive and repository for all ACS Meetings & Expositions abstracts and select posters or presentations.