Machine Learning Used to Formulate Customized Odors to Determine Specific Smell Impression

Current approaches only forecast olfactory impressions from the physiochemical features of odorants, although that particular approach cannot predict the sensing data, which is indispensable for creating smells.

A Nanowerk report specified that the sense of smell is one of the "basic senses of animal species." More so, it is crucial to look for food, sense danger, and realize attraction.

Humans can detect smells, or odorants, with olfactory receptors expressed in olfactory nerve cells. Such olfactory impressions of odorants on nerve cells are linked to their molecular features, not to mention physicochemical properties. This makes it possible to tailor odors to create an intended odor impression.

Predicting Molecular Traits Based on Odor Impression

To address this issue, researchers from the Tokyo Institute of Technology, also called Tokyo Tech, have used an innovative approach to solving the inverse problem.

Rather than predicting the molecular data's smell, this approach predicts molecular traits based on the odor impression.

This is attained by utilizing standard mass spectrum data and ML or machine learning models, according to professor Tekamichi Nakamoto, the leader of the study initiative conducted by Tokyo Tech.

He added that they then predicted the mass spectrum from odor impression inversely based on the formerly developed forward model.

An Essential Part of the Recipe for New Preparation of Odors

The study Exploration of Sensing Data to Realize Intended Odor Impression Using Mass Spectrum to Odor Mixture has been published in the journal PLoS One.

Essentially, the mass spectra of odor mixtures are attained by a linear combination of the mass spectra of single components.

This simple method enables the rapid preparation of spectra of odor mixtures and can predict the needed mixing ratio, as an essential part of the recipe for new preparation of odors.

For instance, explained the researchers, they demonstrate which molecules give the mass spectrum of apple flavor with enhanced sweet and fruit impressions.

Tailor-Made Odors

Professor Nakamoto explained that combinations of either 59 or 60 molecules offer the same mass spectrum as the one attained from the indicated odor impression.

The professor also highlighted that they could theoretically prepare the direct scent with this information, not to mention the proper mixing ratio required for a particular impression.

Such a novel method detailed in this research can offer exact predictions of the physicochemical properties of odor mixtures and the mixing ratios needed for their preparation, opening the door to endless tailor-made odors.

Olfactory Sense

ScienceDirect defines the olfactory sense as "one of the five human senses," which include weight, hearing, taste, touch, and smell or olfactory.

A study associated with sight and hearing is a common field of engineering. Nonetheless, studies about smell or taste are less popular.

One of the reasons such senses are investigated less is the difficulty of analyzing smell and taste, as every individual's subjective impressions decide these evaluations.

More so, such senses cannot be gauged objectively. For instance, one should consider the difficult case of deciding on the fragrance of a new perfume. This decision, in particular, needs that the perfume designers have common perceptions about perfume.

Related information about the use of machine learning in the sense of smell is shown on Changelog's YouTube video below:

Check out more news and information on Machine Learning in Science Times.