Molecular Odor Prediction Using Olfactory Receptor Information.

Area of Science:

• Computational chemistry
• Chemosensation
• Molecular modeling

Background:

• Fragrance development faces significant labor, cost, and human resource challenges.
• Predicting molecular odor from structure alone has practical limitations.
• Understanding human olfaction mechanisms is key to odor prediction.

Purpose of the Study:

• To develop a predictive model for molecular odor, particularly for molecules with subtle structural differences.
• To overcome limitations in current odor prediction methods for fragrance development.
• To leverage human olfaction mechanisms for improved odor prediction.

Main Methods:

• Developed three models based on a three-level analysis of human olfaction.
• Constructed a classification model for molecule-olfactory receptor binding.
• Developed a regression model for binding strength and a classification model for odor presence based on binding.
• Utilized olfactory receptors as descriptors to differentiate similar molecular odors.

Main Results:

• Successfully predicted odor differences between structurally similar molecules.
• Demonstrated accurate prediction for optical isomers, which often have similar structures but distinct odors.
• The multi-level modeling approach effectively captured nuances in odor perception.

Conclusions:

• The developed models offer a promising solution for predicting molecular odors, especially for challenging cases like optical isomers.
• This approach can significantly streamline the fragrance development framework.
• Further refinement of the models holds potential for broader applications in chemical and sensory sciences.