Drug-induced cardiotoxicity or torsades de pointes (TdP), a potentially lethal cardiac ventricular arrhythmia, is an adverse effect that has long been a leading cause of attrition during drug development.
Minimizing the risk of this cardiotoxic effect is thus an important task during the drug development process and regulatory guidelines require new drugs to be evaluated for pro-arrhytmic risk before entering clinical testing. At present, block of the cardiac potassium channel hERG and human QT intervals are assessed as part of the current safety guidelines. Although a block of the cardiac potassium channel hERG and subsequent prolongation of the cardiac QT interval are common features of cardiotoxic drugs, there is no simple one-to-one correlation. TdP involves changes in cardiac cell repolarisation, which is dependent on the concerted activity of several ion channels including hERG, Na-, and Ca-channels. Too much emphasis on hERG as a marker has most likely hampered the development of new drugs by premature discontinuation from development.
We aim to directly assess the primary clinical endpoint, namely ventricular proarrhythmia (i.e., cardiotoxicity). To achieve this, we use a data driven approach based on published data to train a neural network architecture.
The technology is made easily accessible to potential users via a web based demonstrator.