Integrated QSAR Models for Prediction of Serotonergic Activity: Machine Learning Unveiling Activity and Selectivity...

Area of Science:

• Computational Chemistry
• Pharmacology
• Artificial Intelligence

Background:

• Predicting compound activity and selectivity for serotonin receptors is crucial for drug discovery.
• Current methods are limited, often focusing on receptor structure or a few targets.
• Identifying key ligand features can minimize adverse events and maximize therapeutic efficacy.

Purpose of the Study:

• To develop AI-based models for predicting serotonergic activity and selectivity.
• To identify molecular descriptors that determine ligand affinity for serotonin receptors.
• To provide insights into structure-activity relationships for drug design.

Main Methods:

• Utilized Automated Machine Learning (Mljar) to build predictive models.
• Employed SHAP importance analysis to interpret model predictions.
• Focused on molecular descriptors to represent ligands for AI analysis.

Main Results:

• Developed highly efficient AI models for predicting serotonergic activity and selectivity.
• Identified key molecular features influencing ligand affinity and receptor subtype selectivity.
• Clarified the relationship between specific molecular descriptors and their impact on activity.

Conclusions:

• AI models can effectively predict compound serotonergic activity and selectivity.
• Key ligand features influencing receptor interactions have been highlighted.
• The findings support the development of safer and more effective serotonergic drugs.