Brian K. Lee and colleagues have developed a Principal Odor Map (POM) that models the connections between an odorant’s chemical structure with its perceptual property of smell. The map performed as well as some highly trained human “sniffers” in describing odor quality, and could be used for predicting odor intensity and perceptual similarity between odorants. The map moves researchers closer to being able to match molecular properties of odorants to their perceptual properties, a challenge that has proved difficult for olfactory science. (For other senses, neuroscientists have been able to map light wavelengths to color and frequency to pitch, for instance.) Lee et al. used a type of neural network for data processing, called graph neural networks, to develop the POM. One of the things POM could be used for is exploring new odorants; the researchers compiled a list of about 500,000 potential odorants that have never been synthesized, plotting them in the POM to get an idea of how they might smell. Exploring this space would take approximately 70 person-years of continuous smelling time to collect using trained human sniffers. The POM could also aid research on digitizing smells, the authors note.
Science
A principal odor map unifies diverse tasks in olfactory perception
1-Sep-2023