Prediction of the hERG Potassium Channel Inhibition Potential with Use of the Artificial Neural Networks

摘要

The procedure of drug development is complex and time consuming. Informally this procedure is divided into strongly inter-dependent phases and regulated by the regulatory bodies like FDA (Food and Drug Administration). They provide general guidelines for the pharmaceutical industry. Recently more and more effort has been invested in the early toxicity assessment. One of the possible and potentially dangerous effect is a triggered by drugs acquired long QT syndrome (LQTS) which can lead to the fatal ventricular arrhythmia. In the last decades LQTS was responsible for the withdrawal of several drugs from the market. In most drugs known causing TdP (Torsade de Pointes) the QT segment prolongation in the ECG (electrocardiography) results from inhibition of fast potassium channel (encoded as hERG). Therefore early prediction of the hERG channel interaction potential has become a major pharmacological safety concern for the substances being drug candidates. The objective of this research is to develop a reliable empirical model for the potassium channel inhibition prediction, based on the previously published and transparent database. The input data consisted of parameters describing chemical structure and physico-chemical parameters of the substances. Artificial neural networks were chosen as the algorithms for the models development. Classifiers were built on the training set containing 447 records. Two test modes were applied to the model performance assessment: standard 10-fold cross validation procedure and validation based on the external test set of 45 records. The performance of the best model estimated in 10-fold CV was 76% (78% for positive and 74% for negative output respectively). Best obtained model properly predicted 89% instances from the external validation set.