Classifying Drugs by their Arrhythmogenic Risk Using Machine Learning.

Area of Science:

• Cardiovascular Pharmacology
• Computational Biology
• Machine Learning in Drug Safety

Background:

• Current drug safety evaluations for cardiac arrhythmias are inefficient and costly.
• Adverse cardiac effects, particularly arrhythmias, pose significant risks.
• Developing faster, more accurate drug safety assessments is crucial for efficient drug development.

Purpose of the Study:

• To develop a novel risk estimator for stratifying drugs based on their proarrhythmic potential.
• To integrate multiscale experimental and computational data using machine learning for a holistic drug effect assessment.
• To identify key ionic current interactions governing drug-induced arrhythmias.

Main Methods:

• Combined multiscale experiments and simulations with high-performance computing and machine learning.
• Integrated data across 10 orders of magnitude in space and time.
• Utilized Gaussian process classification to develop a predictive model based on ion current block concentrations.

Main Results:

• Drug-induced arrhythmias are primarily influenced by the interplay between the rapid delayed rectifier potassium current and the L-type calcium current.
• A Gaussian process classifier accurately stratified 22 common drugs into safe and arrhythmic categories based on current block concentrations.
• The model elucidates conditions under which L-type calcium current blockade can prevent arrhythmias.

Conclusions:

• Machine learning offers a more accurate and comprehensive mechanistic assessment of drug proarrhythmic potential.
• The developed risk assessment tool can accelerate drug development and the design of safer drugs.
• This approach facilitates the establishment of science-based criteria to reduce heart rhythm disorders.