Google AI Can Tell What Things Smells Like By The Molecular Structure
A Google AI can now tell what things smell like by the molecular structure. Some experts, however, are dubious about the AI’s effectiveness since they argue it does not consider how the human brain processes and interprets scent data.
Google’s computer scientists have created an artificial intelligence (AI) tool that can predict a substance’s smell based on its chemical composition.
It builds on work from 2019 where the technology explained scents utilizing words and employs an “odour map” to visualize the indicative scents of a specific molecule.
On the map, similar-smelling points cluster together, making it possible to anticipate what a substance will smell like before humans give it a sniff.
The scientists from Cambridge, Massachusetts, USA, concluded that “the model is as reliable as a human in describing odor quality.”
They envision using the AI model to find novel fragrances or flavor profiles for food preparation.
It might even be able to recommend new, efficient mosquito repellents or other disease-carrying insect deterrents.
Left: An example of a color map with values for hue, saturation, and wavelength of light. Colors that are similar are close together. Right: The Principal Odour Map. Individual molecules correspond to points (grey), and the locations of these points reflect predictions of their odor character.
It is more challenging to map the variety of smells that can be detected by our noses than, say, the colors that can be seen by our eyes.
This is due to the fact that while we have more than 300 equivalent scent receptors, our eyes’ cone sensors can only detect red, blue, and green colors.
This indicates that there are a wide variety of smells that a person can detect and even more smells that a person may be able to detect.
Since there are not any distinguishing scents that are known to smell the same to everyone, our perceptions of what things smell like are also subjective.
The Google team used flavour and fragrance datasets from over 5,000 different molecules to train a neural network, which resulted in the new “Principal Odour Map” (POM).
The researchers described three experiments they ran to determine the Principal Odour Map’s applicability in a paper that was published this month in bioRxiv.
They asked a panel of 15 experts to describe the aroma of 320 molecules that the AI had not been trained on in order to gauge its precision.
Since each person’s perception of an odor varied slightly, the results of the AI for these molecules were compared to the average of all the panelists.
“We found that the predictions of the model were closer to the consensus than the average panelist was,” the researchers wrote in a Google blog post.
“In other words, the model demonstrated an exceptional ability to predict odor from a molecule’s structure.”
Top: Prediction of scent profiles by Google’s AI model for the molecule 2,3-dihydrobenzofuran-5-carboxaldehyde. Bottom: Average ratings given by trained panellists for the same molecule. Each bar corresponds to one odour character label. The model correctly identifies four of the top five, with high confidence. R = Prediction’s correlation to the full set of 55 labels
The strength of the fragrance, its resemblance to other odors, and how other animals would perceive it were all precisely detected by the AI.
The researchers said: “We found that the map could successfully predict the activity of sensory receptors, neurons, and behavior in most animals that olfactory neuroscientists have studied, including mice and insects.”
For the latter, they gathered information on how different species interpret molecules representing ‘metabolic states’ – or metabolites – such as ripe or rotten, nourishing or inactive, and healthy or sick.
They discovered that if a long series of metabolic reactions is necessary to convert one metabolite to another, the two will appear quite far apart on the map.
Odorous metabolites that are highly similar and appear near together, on the other hand, simply require a few metabolic events to be transformed into one another.
This is consistent with the evolutionary idea that animals’ capacity to smell aids them in distinguishing between various metabolic states.
“The POM shows that olfaction is linked to our natural world through the structure of metabolism and, perhaps surprisingly, captures fundamental principles of biology,” said the researchers.
It is envisaged that by using this data, the model will be able to identify both human and animal diseases.
The final test was designed to see whether the team’s AI could recognize compounds that might function as insect repellents.
Using two datasets that demonstrate how well a particular molecule might deter mosquitoes, they retrained the neural network.
Left: Illustration visualising metabolic reactions found in 17 species across 4 kingdoms. Each circle is a distinct metabolite molecule and an arrow indicates that there is a metabolic reaction that converts one molecule to another. Metabolites without an odour are shown in grey. The metabolic distance between two odorous metabolites is the minimum number of reactions necessary to convert one into the other. In the path shown in bold, the distance is 3. Right: Metabolic distance was highly correlated with distance in the POM
It was discovered to be capable of predicting the mosquito-repellency of almost any chemical, including those that were not included in the datasets that had undergone experimental validation.
The researchers wrote: “We…found over a dozen of them with repellency at least as high as DEET, the active ingredient in most insect repellents.”
“Less expensive, longer lasting, and safer repellents can reduce the worldwide incidence of diseases like malaria, potentially saving countless lives.”
The same approach could be used in the future to identify chemicals that deter other pathogen-carrying species.
The researchers retrained the neural network using two datasets that describe how well a given molecule can keep mosquitoes away. One was from the USDA, and the other used laboratory mosquito feeder assay data, which was used iteratively to improve predictions.
The model identified many molecules that showed repellency greater than the most common repellents used today (DEET and picaridin)
Some experts, however, are dubious about the AI’s effectiveness since they argue it does not consider how the human brain processes and interprets scent data.
Additionally, the work does not take into consideration smells that are produced by intricate arrangements of different scent molecules.
Barry Smith, from the School of Advanced Stud at the University of London, told New Scientist: “Nearly all of the smells we are aware of – wine, coffee, soap, other people, the sea – are due to a mixture of several hundred volatile molecules.”
“Eating food, there’s the saliva in our mouths, there’s the taste receptors contributing, the texture of the food.”
“Many things are interacting to give you a multi-sensory experience. So I think we are still far away from simply predicting flavour from food molecules.”
“We will still have to fill in the biology eventually if we want to understand how humans perceive odours.”
- Source : GreatGameIndia
