TSSF-hERG: A machine-learning-based hERG potassium channel-specific scoring function for chemical cardiotoxicity...

Area of Science:

• Computational chemistry
• Pharmacology
• Cardiovascular research

Background:

• The human ether-à-go-go-related gene (hERG) channel is crucial for cardiomyocyte repolarization.
• hERG channel blockers can cause cardiotoxicity, leading to drug candidate disqualification.
• In silico prediction of drug-hERG binding affinity can accelerate drug discovery.

Purpose of the Study:

• To develop and validate a machine learning model for predicting drug binding affinity to the hERG channel.
• To improve the accuracy of in silico cardiotoxicity assessments.
• To reduce the time and cost associated with experimental drug screening.

Main Methods:

• Collected 9,215 compounds with AutoDock Vina docking structures for training.
• Utilized five machine learning algorithms combined with ligand and interaction features.
• Employed tenfold cross-validation and external verification for model assessment.

Main Results:

• The Support Vector Regression (SVR) model, TSSF-hERG, demonstrated superior performance.
• TSSF-hERG achieved a Pearson's correlation coefficient (Rp) of 0.765 and Spearman's rank correlation coefficient (Rs) of 0.757.
• The model outperformed AutoDock Vina's scoring function and RF-Score.

Conclusions:

• TSSF-hERG significantly enhances the prediction of drug-hERG binding affinity.
• The model can be utilized for virtual screening of drug candidates to predict hERG-related cardiotoxicity.
• This approach aids in the early identification of potential cardiotoxic compounds in drug development.