A Structure-Based Approach for Predicting Odor Similarity of Molecules via Docking Simulations with Human Olfactory...

Area of Science:

• Computational chemistry
• Chemosensory science
• Molecular modeling

Background:

• Human odor recognition mechanisms are poorly understood, hindering scent prediction from molecular structure.
• Odor perception is complex and lacks standardized labels, making absolute classification challenging.
• Current methods often rely on molecular structure similarity, which doesn't fully capture olfactory perception.

Purpose of the Study:

• To develop a relative odor prediction framework based on odor similarity.
• To establish a structure-based, receptor-level approach for computational olfaction.
• To move beyond traditional QSAR methods for scent prediction.

Main Methods:

• Constructed 3D structures of ~400 human olfactory receptors (hORs) using AlphaFold2.
• Performed molecular docking simulations between hORs and odorant compounds.
• Represented odorants as docking score vectors to infer odor similarity statistically.

Main Results:

• Demonstrated that similar docking profiles correlate with similar olfactory perceptions.
• Successfully predicted relative odor similarity between molecules, including novel compounds.
• Validated the approach using odorant molecules from the ATLAS database.

Conclusions:

• The proposed docking-based method provides a reliable framework for predicting relative odor similarity.
• This receptor-level approach offers a significant advancement over traditional QSAR methods.
• The study lays the groundwork for a more accurate computational understanding of olfaction.