Google’s AI used artificial intelligence to create a map that associates odors with molecular structures. It is as reliable as humans when it comes to describing the smell of a substance, and the researchers behind this work say that it is an important step toward digitizing the smell.
A Little About the Model
Mapping the molecular structure for odor recognition is an important challenge in olfactory functioning. Here, Google AI researchers use a graph neural network (GNN) to create a master map of odors (POM) to maintain perceptual relationships and help predict odor quality for substances. new scent. The model is as reliable as humans in describing odor quality: across a potential validation set of 400 new odorants, the odor profiles produced by the model are more in line with the table averages. training (n = 15) against the panel average. By applying simple, interpretable, and theoretically rooted transformations, the POM outperformed the valence models in several other odor prediction tasks, showing that the POM code successfully generated a general map of structure-odor relationships. This approach largely enables odor prediction and paves the way for odor digitization.
Odor is produced by molecules floating in the air, entering our nose, and binding to sensory receptors. There are potentially billions of molecules that can produce odors, so it is difficult to categorize or predict which types produce which odors. Sensory maps can help us with this problem. Color vision features the most familiar examples of these maps, from the color wheel the team learned in elementary school to the more complex variations used to perform color correction in video production. Although these maps have existed for centuries, useful maps for odors are still lacking because smell is a harder problem to crack, and molecules differ in far more ways than photons; Data collection requires physical proximity between smell and smell, and the human eye has only three color-sensitive receptors while the human nose has >300 for the smell. As a result, previous attempts to create odor maps have not gained traction.
In 2019, Google AI developed a graph neural network (GNN) model that began to explore thousands of examples of distinct molecules paired with the smell labels that they evoke, e.g., “beefy”, “floral”, or “minty”, to learn the relationship between a molecule’s structure and the probability that such a molecule would have each smell label. The embedding space of this model contains a representation of each molecule as a fixed-length vector describing that molecule in terms of its odor, much as the RGB value of a visual stimulus describes its color.
This time, the team introduced the “Principal Odor Map” (POM), which identifies the vector representation of each odorous molecule in the model’s embedding space as a single point in a high-dimensional space. The POM has the properties of a sensory map: first, pairs of perceptually similar odors correspond to two nearby points in the POM (by analogy, red is nearer to orange than to green on the color wheel). Second, the POM enables us to predict and discover new odors and the molecules that produce them.
Google AI discovered that their modeling approach to smell prediction could be used to draw a Principal Odor Map for tackling odor-related problems more generally. This map was the key to measuring smell: it answered a range of questions about novel smells and the molecules that produce them, it connected smells back to their origins in evolution and the natural world, and it is helping us tackle important human-health challenges that affect millions of people. Going forward, it is hoped that this approach can be used to find new solutions to problems in food and fragrance formulation, environmental quality monitoring, and the detection of human and animal diseases.
To Read More IT Related Articles Click Here