Machine learning vs. field 3D-QSAR models for serotonin 2A receptor psychoactive substances identification

Area of Science:

• Medicinal Chemistry
• Computational Chemistry
• Pharmacology

Background:

• Serotonergic psychedelics acting on the serotonin 2A receptor (5HT2AR) are a significant part of new psychoactive substances (NPS).
• Accurate prediction of ligand affinity is crucial for identifying and designing novel compounds targeting 5HT2AR.

Purpose of the Study:

• To develop and validate quantitative structure-activity relationship (QSAR) models for predicting 5HT2AR ligand affinity.
• To utilize these models for exploring natural products, classifying NPS, and designing new potential 5HT2AR ligands.

Main Methods:

• Development of five QSAR models using Forge software with 375 molecules.
• Statistical analysis to confirm model quality and predictive capabilities.
• Combination of machine learning and 3D-QSAR models into a consensus model, incorporating a pharmacophore filter.

Main Results:

• Successfully developed robust QSAR models with strong predictive and descriptive capabilities for 5HT2AR ligands.
• The consensus model and pharmacophore filter were applied to analyze a large dataset of natural products and classify recent NPS.
• Identified potential new active molecules targeting the 5HT2AR.

Conclusions:

• The developed QSAR models provide an effective tool for the investigation and identification of unclassified 5HT2AR ligands.
• This approach facilitates the classification of emerging NPS and aids in the rational design of novel psychoactive compounds.
• The study supports the identification and classification of new 5HT2AR ligands, including NPS.