The smell of an object is the only type of information that goes directly from the sensory organ. Unlike the other kinds of sensory input, which pass first through the different regions of the brain, the sensory information from the nose travels directly to the memory and emotional centers of the brain. This direct route is also why odors can trigger specific and intense memories.
To explore the connection between the structure and odor of a chemical, a team of researchers designed a type of artificial intelligence (AI) system that can describe the smell of a compound just by analyzing its molecular structure.
Predicting the Odor Profile of a Molecule
Vision and hearing have well-developed maps that relate physical properties such as wavelength and frequency to perceive properties like color of light and pitch of sound. These properties can also be measured and assessed by instruments. When it comes to olfaction, there is currently no way to accurately measure and predict the odor of a molecule based on its structure.
For this reason, a team of researchers from start-up company Osmo in Cambridge, Massachusetts, used an artificial intelligence system to develop a principal odor map (POM), which can preserve perceptual relationships and allow odor quality prediction for previously uncharacterized odorants. This type of AI system is called a neural network and works by assigning descriptive words to an odorant.
The new machine learning-generated model correctly predicts the scent of the numerous exceptions where the smell and structure do not match. It can effectively recognize molecules that look different but smell the same. Similarly, it can identify molecules that appear similar but have different smells.
Experts from the University of Reading assessed the samples' purity in testing the AI. They used gas chromatography to separate the trace levels of impurities and the target molecules. As the particles were eluted one by one from the instrument, the experts could smell the individual molecules and determine whether the scent of any trace compounds was masking the odor of the target molecule.
Once the AI had been trained with this data, it showed an excellent ability to predict a new compound's smell, even matching the average scent scores of a group of humans.
READ ALSO: Decoding the Works of Smell Receptors: Scientists Determine How Scent Molecules Are Captured
Uses of Principal Odor Map
The odor map created by the researchers can play a significant role in the work of synthetic chemists in the food and fragrance industries. It can provide valuable insights into producing more sustainable flavors and fragrances, opening up an untapped source of thousands or millions of potential odorants.
According to Professor Jane Parker from the University of Reading, the principal odor map does not only work for known odorants and those that have very similar structures. This tool can also describe a wide subset of unrelated molecules. Furthermore, it can serve as a guide in designing new synthetic fragrances and providing insights into how the human brain interprets smell.
RELATED ARTICLE: Consistent Smell of Objects Regardless of External Factors Explained Through Neural Response in Locusts
Check out more news and information on Odor in Science Times.
