Machine-learning technique, QSAR and molecular dynamics for hERG-drug interactions

Area of Science:

• Computational chemistry and cheminformatics
• Pharmacology and toxicology
• Drug discovery and development

Background:

• The human Ether-à-go-go-Related Gene (hERG) channel is a critical anti-target in medication cardiotoxicity due to its role in cardiac repolarization.
• hERG channel inhibition can lead to QT prolongation, Torsades de Pointes, and sudden cardiac death.
• Traditional experimental screening for hERG liability is costly, time-consuming, and complex.

Purpose of the Study:

• To develop robust quantitative structure-activity relationship (QSAR) and predictive classification models for KCNH2 liability.
• To create computational (in silico) tools for early identification of potential hERG blockers in drug discovery.
• To investigate the molecular mechanisms of hERG-related cardiotoxicity using computational methods.

Main Methods:

• Utilized a curated dataset of 6766 compounds with 2D descriptors.
• Employed various machine learning algorithms including Decision Tree, Random Forest, Logistic Regression, Ada Boosting, kNN, SVM, Naïve Bayes, neural networks, and stochastic gradient classification.
• Performed molecular docking and 200 ns molecular dynamics simulations to analyze hERG-ligand interactions.

Main Results:

• Developed descriptor-based QSAR and classification models to predict hERG liability.
• Achieved predictive classification based on IC50 values, categorizing compounds as hERG-positive (blockers) or hERG-negative (non-blockers).
• Identified key interacting hERG residues (LEU:697, THR:708, PHE:656, HIS:674, HIS:703, TRP:705, ASN:709) and confirmed stable hERG-ligand complex behavior through simulations.

Conclusions:

• Machine learning models effectively predict hERG channel liability using 2D molecular descriptors.
• In silico approaches can significantly aid in the early-stage identification of potential cardiotoxic compounds.
• Computational modeling provides valuable insights into the molecular basis of hERG-related cardiotoxicity, facilitating safer drug development.