What does Covid-19 tell us about the mystery of odors?

Meyer began to feel as if he knew people personally — those who described the smell in terms of tea and fruit, or meat and petrol, or blue powerade and lollipop. The way they explained their senses felt very intimate, he later said, “you could almost see what type of person they were.” He is often asked to do so in the lab, sniffing out a single isolated molecule (when the more familiar smell of coffee is a blend of hundreds of them) and explaining the smell. I was beginning to be convinced that people believed they were not good at it. Apart from their real life and the scent context that is really important to them. “People are very, very verbal,” he said, given the right opportunities.

This was exciting news for IBM researcher Meyer, who specializes in analyzing biological data using algorithms and insisted that GCCR studies should include open text boxes. For years, olfactory scientists have processed just a few very inadequate datasets that link different chemicals to the way humans perceive them. For example, in the late 1960s, there was a record created by a perfumer who explained thousands of scents, and he continued his research based on a single “Atlas of Odor Character Profiles” published in 1985. It was. This is to sniff, evaluate and name various single molecules and chemical mixtures according to the provided list of descriptors that many scientists felt were flawed and dated. Volunteer sought after.

More recently, Meyer and many others have used a new dataset carefully created by scientists at Rockefeller University in New York and published in 2016 (Leslie Voshall and her colleagues). I visited the lab in 2014 while creating the data. I was surprised to be able to “smell” one of the vials, but it probably only caused my trigeminal nervous system. When she told Voshall that it was like mint, she replied: However, although the new dataset was significantly improved, 55 people sniffed 480 different molecules, and the intensity, Evaluated by comfort, friendliness, consistency with a list of 20 descriptions such as “garlic”, “spices”, “flowers”, etc. “bakery”, “muskey”, “urine” etc. It showed if it was limited.

This is why Meyer, along with his colleague Guillermo del Toro, requested these open text boxes in a GCCR study. They were interested in the potential of natural language processing, a branch of machine learning that uses algorithms to analyze patterns of human expression. Chekki has already used this technique to analyze the details of how people speak and predict the early onset of Alzheimer’s disease at the most treatable time. Many researchers have written about the possibility of using artificial intelligence to ultimately create predictive maps of olfaction. We also investigated the association of olfactory changes with all the diseases associated with those changes, but did not provide sufficient data.

Covid is now providing researchers with large and complex datasets that link their olfactory experience with the progression of certain illnesses. It was not constrained by numerical rankings, single molecules, or some provided adjectives, but instead people are free to talk about the real scent in all of the complex and subjective glory in the real world. Made it possible.

When Meyer and Cecchi’s colleague Raquel Norel finish analyzing free-form responses from English-speaking respondents, their text analysis predicts a Covid diagnosis similar to a numerical assessment of people’s odor loss. I was surprised and delighted to be a thing. The algorithm worked because people with Covid talk about odors in very different languages ​​than people without odors. Even people who have not completely lost their sense of smell can feel the same by repeating words such as “metal,” “rot,” “chemistry,” “acid,” “sour,” “burnt,” and “urine.” I tended to explain. .. It’s an encouraging finding that we couldn’t wait to explore deeper, first with GCCR responses in other languages, and then with other datasets related to other diseases in the future. It was a proof of concept. Meyer was excited when he talked about it. “Anything that changes the smell,” he told me. “Depression, schizophrenia, Alzheimer’s disease, Parkinson’s disease, neurodegeneration, cognitive and neuropsychiatric disorders. As they say, the whole Enchillada.”

I had a hard time Imagine the olfactory “map” that scientists have long dreamed of. Is it like the periodic table I asked the mainland? He suggested instead of a map created by scientists in a “color space” where colors are placed to show mathematical relationships and mixes, I think. “Until people started inventing color TV, Photoshop, etc., we didn’t know how useful color space was,” he explained, and the map itself is not the goal, but what we smell using it. We do so, adding that it is the ability to understand what to sniff. After that, what’s really interesting is the application we can’t imagine yet. “It’s hard to understand the usefulness of a map until it’s available,” he said.