Scientists use machine learning to predict odors based on worm brain activity-ScienceDaily

“When we started investigating the effects of these sensory stimuli on the connections of individual cells and worms in the brain, we found some unexpected things,” said the new work, a member of the Institute for Molecular Neurobiology. Senior author Chalasani has published in the journal. PLOS Computational Biology November 9, 2021.

Chalasani is interested in how the brain processes information from the outside world at the cellular level. Researchers cannot simultaneously track the activity of each of the 86 billion living human brain cells, but microscopic worms can. Caenorhabditis elegans, 302 There are only neurons. Chalasani describes it with a simple animal like this C. Elegance, Researchers can monitor individual neurons as animals perform actions. That level of resolution is currently not possible even with humans and mice.

Chalasani’s team set out to study how C. Elegance Neurons respond to the odors of each of five different chemicals: benzaldehyde, diacetyl, isoamyl alcohol, 2-nonanone, and sodium chloride.Previous research C. Elegance You can distinguish between these chemicals. To humans, these chemicals smell like almonds, buttered popcorn, bananas, cheese, and salt. Researchers know the identities of the few sensory neurons that directly sense these stimuli, but Chalasani’s group was more interested in how the rest of the brain reacts.

Designed researcher C. Elegance Therefore, each of the 302 neurons contained a fluorescent sensor that would light up when the neuron was active. Next, we observed under a microscope how 48 types of worms were repeatedly exposed to bursts of 5 chemicals. On average, 50 or 60 neurons were activated in response to each chemical.

By examining the basic properties of the dataset, such as the number of cells that were active at each point in time, Chalasani and his colleagues could not immediately distinguish between the different chemicals. So they turned to a mathematical approach called graph theory. It analyzes the collective interactions between pairs of cells. When one cell is activated, how does the activity of the other cells change accordingly?

This approach is always C. Elegance When exposed to sodium chloride (salt), there was initially a burst of activity in a set of neurons (probably sensory neurons), but after about 30 seconds, triplets of other neurons began to strongly regulate their activity. I did. These same distinct triplets were not found after other stimuli, and researchers were able to pinpoint when the worm was exposed to salt based solely on brain patterns.

“”C. Elegance It seems to add high value to salt sensing by responding using a completely different circuit configuration in the brain. This may be because salt often represents the bacteria that feed the worm.

Researchers then used machine learning algorithms to identify other, more subtle differences in the brain’s response to each of the five chemicals. The algorithm was able to learn to distinguish between the neural response to salt and benzaldehyde, but often confused the other three chemicals.

“Whatever analysis we do, it’s the beginning, but we still have only a partial answer as to how the brain distinguishes between these things,” says Chalasani.

Still, he points out how the team approached the study-looking at the response of the entire brain network to stimuli, rather than focusing on a small set of sensory neurons and whether they are activated. , Apply graph theory — paving the way for a more complex and holistic study of how the brain responds to stimuli.

Of course, the researcher’s ultimate goal is not to read the mind of a microscopic worm, but how humans encode information in the brain, and this is related to sensory processing disorders and anxiety. A better understanding of what happens when things go wrong, Attention Deficit Hyperactivity Disorder (ADHD), Autism Spectrum Disorders, etc.

Other authors of the new study were Sackett Nablaha of the Cold Spring Harbor Laboratory and Javier Howe of the University of California, San Diego. This work was supported by grants from the Pew Charitable Trust, the National Institutes of Health, and the National Science Foundation.