Initial Development of Automated Machine Learning-Assisted Prediction Tools for Aryl Hydrocarbon Receptor Activators

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• The aryl hydrocarbon receptor (AhR) is vital for immune and metabolic functions.
• Designing potent AhR modulators is challenging due to ligand diversity.
• Novel tools are needed for drug discovery targeting AhR.

Purpose of the Study:

• To develop and compare quantitative structure-activity relationship (QSAR) models for predicting AhR activity.
• To identify the most effective QSAR modeling approach for AhR modulation.
• To provide a foundation for future drug design targeting AhR.

Main Methods:

• Combined ChEMBL and WIPO databases for 978 molecules with EC50 values.
• Developed classification and regression QSAR models using the mljar platform.
• Employed 10-fold cross-validation and SHAP for model interpretation.

Main Results:

• Classification model achieved 0.760 accuracy and 0.789 F1 score.
• Regression model yielded RMSE of 5444 and R² of 0.208.
• Developed an online AhR web application using the best classification model.

Conclusions:

• QSAR models can predict AhR activity, aiding in understanding structure-activity relationships.
• The developed web application serves as a practical tool for researchers.
• Findings support further development of QSAR models for AhR ligand design.