A Computational Pipeline to Predict Cardiotoxicity: From the Atom to the Rhythm

Area of Science:

• Computational biology
• Cardiovascular pharmacology
• Toxicology

Background:

• Drug-induced proarrhythmia is linked to QT interval prolongation, but this marker lacks selectivity, leading to premature drug candidate elimination.
• Existing methods for predicting cardiotoxicity are insufficient, hindering drug discovery and safety evaluations.

Purpose of the Study:

• To develop a predictive model linking drug chemistry to cardiac rhythm impact.
• To establish a computational framework for predicting drug-induced cardiotoxicity from molecular interactions to organism-level effects.

Main Methods:

• Integrated atomistic simulations of drug-ion channel interactions with cell and tissue-scale models.
• Predicted drug-binding affinities and rates to simulate effects on the hERG channel.
• Validated the multiscale model framework using human clinical data.

Main Results:

• The multiscale model successfully predicted drug effects on cardiac rhythm and validated against human data.
• Identified novel mechanistic insights into proarrhythmic drug interactions at the hERG channel and cellular/tissue levels.
• Machine learning identified key parameters for predicting arrhythmia vulnerability.

Conclusions:

• A novel multiscale model framework enables prediction of cardiac electrotoxicity from atomistic to organismal scales.
• This approach offers mechanistic insights and improves prediction accuracy for drug-induced arrhythmia.
• The model has potential applications in drug discovery, safety screening, and regulatory processes.