Probability Based hERG Blocker Classifiers.

Area of Science:

• Pharmacology
• Computational Chemistry
• Toxicology

Background:

• The US Food and Drug Administration mandates in vitro human ether-a-go-go related (hERG) ion channel affinity testing for drug candidates.
• Traditional quantitative structure-activity relationship (QSAR) models often use supervised 'hard point' classification, providing binary 'yes/no' predictions.

Purpose of the Study:

• To develop novel probabilistic-based prediction models for hERG channel inhibition.
• To offer a more nuanced approach than traditional binary classification methods for drug safety assessment.

Main Methods:

• Employed probabilistic-based methods to create prediction models for hERG inhibition.
• Developed models capable of ascertaining the likelihood of compounds blocking hERG ion channels.

Main Results:

• The probabilistic models provide a method to 'ascertain' compound effects on hERG channels.
• Demonstrated the utility of probabilistic models in ranking compounds based on potential cardio-toxicity.

Conclusions:

• Probabilistic-based methods are valuable for predicting hERG inhibition and ranking compounds for cardio-toxicity.
• The approach shows promise for predicting other toxic properties in drug development.