Can we use machine learning methods to predict odor mixture detection data and design new odors? A new study by Tokyo Tech researchers does just that. The new method is sure to have applications in the food, health, beauty and wellness industries, where scents and fragrances are of great interest.
The sense of smell is one of the basic senses of animal species. It is critical to finding food, noticing attraction, and sensing danger. Humans detect odors, or smells, with olfactory receptors expressed on olfactory nerve cells. These olfactory impressions of odors in nerve cells are associated with their molecular characteristics and physicochemical properties. This makes it possible to tailor scents to create a desired scent impression. Current methods only predict olfactory impressions from the physicochemical characteristics of odours. But that method can’t predict the detection data, which is essential for creating odors.
To address this problem, scientists at the Tokyo Institute of Technology (Tokyo Tech) have employed the innovative strategy of solving the inverse problem. Instead of predicting odor from molecular data, this method predicts molecular features based on odor impression. This is achieved using standard mass spectral data and machine learning (ML) models. “We used a machine learning-based predictive odor model that we had previously developed to obtain the odor impression. We then predicted the mass spectrum from the odor impression inversely based on the previously developed forward model,” he explains. Professor Takamichi Nakamoto, leader of the Tokyo Tech research effort. The findings have been published in PLoS One.
The mass spectrum of the odor mixtures is obtained by a linear combination of the mass spectra of the individual components. This simple method allows rapid preparation of predicted spectra of odor mixtures and can also predict the required mixing ratio, an important part of the recipe for the preparation of new odors. “For example, we show which molecules give the mass spectrum of apple flavor with enhanced impressions of ‘fruit’ and ‘sweet’. Our method of analysis shows that combinations of 59 or 60 molecules give the same mass spectrum as obtained of the scent impression. With this information, and the correct mix ratio needed for a given impression, we could theoretically prepare the desired scent,” highlights Professor Nakamoto.
This novel method described in this study can provide highly accurate predictions of the physicochemical properties of scent mixtures, as well as the mixing ratios required to prepare them, opening the door to a myriad of personalized fragrances.
Looks like the future of scent blends smells good!
Reference:
Hasebe D, Alexandre M, Nakamoto T. Exploring detection data to perform desired odor imprinting using odor mixture mass spectrum. PLUS ONE. 2022;17(8):e0273011. do:10.1371/diary.puts.0273011
This article has been republished from the following materials. Note: the material may have been edited for its length and content. For more information, contact the cited source.