Machine Learning Assisted Approach for Finding Novel High Activity Agonists of Human Ectopic Olfactory Receptors

Area of Science:

• Biochemistry
• Computational Biology
• Pharmacology

Background:

• Olfactory receptors (ORs), the largest G protein-coupled receptor (GPCR) superfamily, are crucial for odorant detection and have non-nasal tissue roles.
• Understanding the vast repertoire of olfactory stimuli and their corresponding ORs is key to advancing olfactory system research and therapeutic applications.

Purpose of the Study:

• To expand the known chemical space of broadly tuned human ectopic ORs, specifically OR1A1 and OR2W1, using machine learning (ML).
• To identify novel, high-activity agonists for OR1A1 and OR2W1 through large-scale virtual screening and experimental validation.

Main Methods:

• Developed and optimized a machine learning model, selecting Random Forest (RF) as the best performer for agonist prediction.
• Screened approximately 23 million compounds across five databases using the trained RF classifier.
• Validated top predicted ligands through molecular docking against developed receptor models and subsequent in vitro assays.

Main Results:

• Identified two novel high-activity agonists for the olfactory receptor OR1A1.
• Discovered one novel high-activity agonist for the olfactory receptor OR2W1.
• Demonstrated the effectiveness of ML-based virtual screening combined with docking and experimental validation for OR agonist discovery.

Conclusions:

• Machine learning, particularly the Random Forest algorithm, is a powerful tool for predicting olfactory receptor agonists.
• This study successfully identified novel agonists for OR1A1 and OR2W1, expanding the chemical space for these receptors.
• The findings pave the way for further exploration of OR functions and the development of therapeutic agents targeting the olfactory system.