Automatic modeling of dynamic drug-hERG channel interactions using three voltage protocols and machine learning...

Area of Science:

• Computational chemistry and pharmacology
• Biophysics of ion channels

Background:

• Drug cardiac safety assessment is crucial for new compound development.
• Evaluating the half-maximal blocking concentration (IC50) of potassium human ether-à-go-go related gene (hERG) channels is standard practice.
• Modeling drug-hERG channel binding dynamics can enhance early cardiac safety evaluations.

Purpose of the Study:

• To develop an automated methodology for generating Markovian models of drug-hERG channel interactions.
• To elucidate drug binding and unbinding states and preferential binding states.
• To capture state-dependent binding, affinities, trapping dynamics, and IKr block onset.

Main Methods:

• Utilized 12 Markovian chains to model diverse drug-binding possibilities (any state, simultaneous binding, state preference).
• Employed three specific voltage clamp protocols to differentiate channel conformational states (open, closed, inactivated).
• Developed a computational tool with a classifier and parameter optimizer using linear interpolation, support vector machines, and simplex method.

Main Results:

• The novel methodology automatically generates dynamic drug models using Markov formulations.
• The tool accurately predicted the class of 92.04% of virtual drugs, with a mean model accuracy of 97.53%.
• Dynamic model generation for an IKr blocker takes under an hour on a standard desktop computer.

Conclusions:

• The methodology effectively models and simulates dynamic drug-hERG channel interactions.
• This approach can improve preclinical assessment of proarrhythmic risk for IKr-inhibiting drugs.
• It also aids in evaluating the efficacy of antiarrhythmic IKr blockers.