In silico methods for the prediction of drug-induced cardiotoxicity

Thesis

Unexpected adverse reactions, especially unsafe cardiac effects, are a major concern of pharmaceutical companies that can prompt them to both discontinue drugs currently in development and withdraw drugs already on the market. Therefore, the safety assessment is a key stage of both the drug development process and the current regulatory framework of clinical trials. Given the importance of unforeseen acute electrophysiological effects in precipitating potentially lethal arrhythmias, the current preclinical testing stages of drug development are largely focused on their detection. However, a substantial number of drugs also affect cardiac function on many other levels, including contractility, mitochondria function and cell signalling. A number of in vitro, in vivo and in silico approaches capable of detecting different types of possible cardiovascular side effects have been proposed recently. Among those, human-based computational methods hold a great potential to increase the productivity of drug discovery pipelines, drive a more rational drug design and replace costly animal experiments that have limited translational ability for humans.

Therefore, the goal of this thesis is to propose a computational approach to predict drug-induced cardiotoxicity. A multi-label machine learning classification approach is used to simultaneously predict multiple forms of clinical cardiac side effects and take into account relationships between those forms of toxicity. In the last part of this thesis, the effects of trafficking impairment, as one of the cardiotoxicity mechanisms, are then investigated using simulations of action potential models.

Authors: Mamoshina, P

• Type of award: DPhil
• Level of award: Doctoral
• Awarding institution: University of Oxford
• Language: English
• Keywords: machine learning; pathway analysis; safety pharmacology; cardiac safety; action potential models; simulations
• Subjects: Action potentials (Electrophysiology); Machine learning; Gene expression