Atomevo-odor: A database for understanding olfactory receptor-odorant pairs with multi-artificial intelligence...

Area of Science:

• Olfactory neuroscience
• Computational chemistry
• Bioinformatics

Background:

• Olfactory receptors (ORs) interact with odorant molecules to encode diverse smells.
• Understanding these interactions is key to deciphering olfactory perception.

Purpose of the Study:

• To construct a comprehensive database (Atomevo-Odor) integrating experimental and AI-predicted OR-odorant responses.
• To analyze odorant classification, functional group relationships, and OR recognition mechanisms.
• To develop a predictive model for OR-odorant interactions.

Main Methods:

• Database construction: Atomevo-Odor combining experimental data and AI predictions.
• Odorant classification using graph theory and unsupervised learning.
• Development of a Convolutional Neural Network (CNN) for OR-odorant response prediction.
• Analysis of odorant functional groups and OR recognition patterns.

Main Results:

• The Atomevo-Odor database provides high-quality OR-odorant response data.
• Odorant classification revealed relationships between functional groups and fragrance types.
• A CNN model accurately predicted OR-odorant responses.
• Identification of novel OR-odorant pairs for further investigation.

Conclusions:

• The study enhances understanding of olfactory perception through a novel database and predictive modeling.
• Insights into OR-odorant recognition mechanisms are provided.
• The findings offer guidance for future experimental design in olfaction research.