Drug-induced inhibition of hERG potassium channels is a major obstacle in drug discovery due to the risk of severe cardiac adverse effects. Applying in silico techniques for early identification of potential hERG blockers is an attractive approach, but the usability of available tools is still quite limited. Despite a few relative successes in hERG inhibition modeling, many recently published models suffer from inherent complexity and lack of interpretability, while simple and well-known physicochemical rules remain mostly qualitative. In this study we attempt to overcome these issues by constructing a large and thoroughly curated hERG inhibition database spanning a range of >2500 diverse chemicals, and utilizing these data to classify compounds as hERG blockers/non-blockers solely on the basis of principal physicochemical determinants such as lipophilicity, ionization, aromaticity, molecular size and flexibility. The proposed classification model was built using Gradient Boosting statistical method known for its ability to account for complex nonlinear relationships and low sensitivity to outliers. The model was able to produce correct classification for almost 80% of validation set compounds indicating that the major part of variation in hERG inhibition propensity can be conveyed by general physichochemical trends, in full consistence with broad ligand specificity of hERG channel. In addition to evaluation and visualization of the physicochemical effects, the obtained model can be used as a baseline predictive tool for more detailed analysis, e.g., exploring the potential of discrete structural modifications to further attennuate hERG liability of candidate compounds.

Deeper insight on physicochemical determinants of hERG inhibitor specificity

248th National Meeting (2014)

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