The use of AI to detect smells has a relatively long history, with use cases ranging from perfume to health and safety. The latest example in the field comes via new research from the University of California, Riverside, which highlights how machine learning is used to understand what a chemical smells like.
“We now can use artificial intelligence to predict how any chemical is going to smell to humans,” the researchers say. “Chemicals that are toxic or harsh in, say, flavors, cosmetics, or household products can be replaced with natural, softer, and safer chemicals.”
Humans take advantage of around 400 odour receptors within the nose. Each of these is activated by a set of chemicals, with the full range of receptors able to detect a huge range of chemicals. The question for the researchers was how the receptors contribute to different perceptual qualities, or percepts.
“We tried to model human olfactory percepts using chemical informatics and machine learning,” they explain. “The power of machine learning is that it is able to evaluate a large number of chemical features and learn what makes a chemical smell like, say, a lemon or a rose or something else. The machine learning algorithm can eventually predict how a new chemical will smell even though we may initially not know if it smells like a lemon or a rose.”
Digitizing smell
By being able to predict how chemicals smell digitally, the researchers believe a new way of prioritizing what chemicals can be used in sectors such as fragrance and food can be created.
“It allows us to rapidly find chemicals that have a novel combination of smells,” they explain. “The technology can help us discover new chemicals that could replace existing ones that are becoming rare, for example, or which are very expensive. It gives us a vast palette of compounds that we can mix and match for any olfactory application. For example, you can now make a mosquito repellent that works on mosquitoes but is pleasant smelling to humans.”
The project began by developing a method to digitally learn the chemical features that activate odorant receptors. The team then screened around 500,000 compounds for new ligands, which are molecules that bind to receptors, for 34 different odorant receptors, before deciding whether the algorithm could predict diverse perceptual qualities of odorants as well as estimating odorant receptor activity.
“Computers might help us better understand human perceptual coding, which appears, in part, to be based on combinations of differently activated ORs,” the researchers explain. “We used hundreds of chemicals that human volunteers previously evaluated, selected ORs that best predicted percepts on a portion of chemicals, and tested that these ORs were also predictive of new chemicals.”
The system was able to successfully predict 146 different percepts, with a surprisingly small number of ORs needed to predict many of these percepts. When this process was tested in fruit flies, similar phenomenon were observed, which the team believe makes predictions easier for the computer as it requires less information.
“Our digital approach using machine learning could open up many opportunities in the food, flavor, and fragrance industries,” the researchers conclude. “We now have an unprecedented ability to find ligands and new flavors and fragrances. Using our computational approach, we can intelligently design volatile chemicals that smell desirable for use and also predict ligands for the 34 human ORs.”
