Jul 29 2020
With the help of machine learning, two scientists from the University of California, Riverside (UCR) have effectively interpreted the smell of chemicals—a breakthrough study that could prove useful in the fragrance and food flavor sectors.
Anandasankar Ray. Image Credit: University of California, Riverside.
We now can use artificial intelligence to predict how any chemical is going to smell to humans. Chemicals that are toxic or harsh in, say, flavors, cosmetics, or household products can be replaced with natural, softer, and safer chemicals.
Anandasankar Ray, Study Senior Author and Professor, Department of Molecular, Cell and Systems Biology, University of California, Riverside
The study was published in the iScience journal.
Odors are perceived by humans when some of the almost 400 odorant receptors (ORs) are stimulated in their noses. A special set of chemicals stimulate each OR. Collectively, the vast OR family is capable of detecting a large chemical space. A crucial question in the area of olfaction is the contribution of the receptors in different percepts or perceptual qualities.
We tried to model human olfactory percepts using chemical informatics and machine learning. 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.
Anandasankar Ray, Study Senior Author and Professor, Department of Molecular, Cell and Systems Biology, University of California, Riverside
Ray believes that a digital approach to predict the smell of chemicals can provide a new way to scientifically prioritize the kinds of chemicals that can be applied in the food, fragrance, and flavor sectors.
“It allows us to rapidly find chemicals that have a novel combination of smells,” Ray added. “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.”
He continued, “For example, you can now make a mosquito repellent that works on mosquitoes but is pleasant smelling to humans.”
The scientists initially created a technique for a computer to learn chemical traits that stimulate familiar human ORs. They subsequently screened about half a million compounds for novel ligands—that is, molecules that adhere to receptors—for a total of 34 ORs. The team then focused on whether the machine learning algorithm that could predict the OR activity could also estimate different perceptual qualities of odorants.
Computers might help us better understand human perceptual coding, which appears, in part, to be based on combinations of differently activated ORs 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.
Joel Kowalewski, Study First Author and Student, Neuroscience Graduate Program, University of California, Riverside
Kowalewski is currently working with Ray.
Both Ray and Kowalewski demonstrated that OR activity had effectively estimated a total of 146 different percepts of chemicals. To the team’s amazement, only a few and not all ORs were required to estimate some of these percepts.
Since the researchers were not able to record the activity from humans’ sensory neurons, they further tested this in the fruit fly called Drosophila melanogaster, and noted an analogous result when estimating the aversion or attraction of the fly to different odorants.
“If predictions are successful with less information, the task of decoding odor perception would then become easier for a computer,” added Kowalewski.
Ray further elaborated that several items, which are available to consumers, include volatile chemicals to render themselves appealing. Around 80% of what is believed to be flavor in food items actually comes from the odors that have an impact on smell.
Fragrances used for perfuming cosmetics and various household goods, including cleaning products, play a crucial role in consumer behavior.
“Our digital approach using machine learning could open up many opportunities in the food, flavor, and fragrance industries. 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,” Ray concluded.
The National Science Foundation and UCR have partly funded the study.
The novel technology has been revealed to the UCR Office of Technology Partnerships, assigned UC case number 2019-131, and is pending patent. It is titled “Methods for identifying, compounds identified and compositions thereof,” and has been licensed to Sensorygen Inc., a startup company.
Established by Ray in 2015, Sensorygen uses artificial intelligence (AI) and computational biology to identify natural replacements for harsh and toxic chemicals that are used in daily products, including identification of new flavors as well as insect repellents.
The study is titled “Predicting human olfactory perception from activities of odorant receptors.”
Journal Reference:
Kowalewski, J., et al. (2020) Predicting human olfactory perception from activities of odorant receptors. iScience. doi.org/10.1016/j.isci.2020.101361.