Tuesday, June 20, 2023


This discovery provides a great example of how machine learning and optogenetics are blowing open our doors of perception.

In this case, scientists created an artificial olfactory receptor (this one derived from OR5A2). You can think of it like the ultimate musk receptor, because after matching it against 100 mammal-nose-brain gene sequences, it's the best-fit for all the animals at once.

But it doesn't really exist in any one animal; it's simply the most in-between of all of them. They call it a "consensus protein". I call it a frankenstein receptor. 

That was machine learning to the rescue, but then they called their friend optogenetics -- they further engineer this artificial protein to produce light when activated. This is a common technique these days that allows us to measure the receptor activity; it's like being able to ask an animal to tell you whether it smells something or not. Then they go back and find all the odorants that match this new frankenstein receptor -- if it lights up, it's a match.

They found no new musks actually, which suggests we know all of them already, but this could work for other odors:

Chemists propose unifying theory of musk - Engineered olfactory receptor may explain why structurally diverse molecules smell similar.
Chemical and Engineering News, Nov 2022

The three receptors known to recognize musk compounds only respond to a subset of musk-scented compounds.

The researchers compared the amino acid sequences for a given odorant receptor across 112 mammal species to determine the most common amino acid at each position and made a receptor with this so-called consensus sequence.

The engineered protein differs from human OR5A2 at 25 of its 324 amino acids ... .

Using the structures of compounds that do and do not activate the receptor, the researchers developed a machine learning model and used it to screen a database of odorant structures and human perceptions. The model, Mainland says, claims to identify known musk molecules much better than prior models trained only using the database. Although they do not report any new musky compounds in the study, the authors say that the Kao Corporation has filed patents related to the work.

via Duke University and Kao Corporation, Tokyo: Proc. Natl. Acad. Sci. U.S.A. 2019, DOI: 10.1073/pnas.1804106115

And this is a pretty big deal in smell science:
First molecular images of olfaction open door to creating novel smells
Mar 2023, phys.org

First molecular-level, 3D picture of how an odor molecule activates a human odorant receptor.

Odorant receptors are notoriously challenging, some say impossible, to make in the lab for such purposes. The Manglik and Matsunami teams looked for one that was abundant in both the body and the nose, thinking it might be easier to make artificially, and one that also could detect water-soluble odorants. They settled on a receptor called OR51E2, which is known to respond to propionate—a molecule that contributes to the pungent smell of Swiss cheese.

This molecular snapshot showed that propionate sticks tightly to OR51E2 thanks to a very specific fit between odorant and receptor. The finding jibes with one of the duties of the olfactory system as a sentinel for danger.

"This receptor is laser focused on trying to sense propionate and may have evolved to help detect when food has gone bad," said Manglik. Receptors for pleasing smells like menthol or caraway might instead interact more loosely with odorants, he speculated.

"We've dreamed of tackling this problem for years," he said. "We now have our first toehold, the first glimpse of how the molecules of smell bind to our odorant receptors. For us, this is just the beginning."

via University of California, San Francisco: Aashish Manglik, Structural basis of odorant recognition by a human odorant receptor, Nature (2023). DOI: 10.1038/s41586-023-05798-y

No comments:

Post a Comment