Determination of hERG channel blockers using a decision tree.

A decision tree approach for the in silico prediction of Torsade de Pointes (TdP)-causing drugs is presented. As TdP is frequently associated with QT-interval prolongation due to inhibition of the rapid activating delayed rectifier potassium channel in the heart (hERG channel), the properties of such blockers were investigated by molecular modeling and semi-empirical AM1 molecular orbital calculations. In addition, we derived a pharmacophoric SMARTS string using structural information from high affinity compounds. A corresponding search in the PubChem database identified several compounds that exhibit QT-interval prolonging activity that were not among our data set. This SMARTS string furthermore showed to be the most significant descriptor in the decision tree approach from which guidelines for the design of safe compounds are suggested.