Fly Study Reveals Neurons That Assist a Discerning Sense of Smell
A well-trained nose can sniff out the smallest details in a scent. A new study may explain this phenomenon.
Complete the form below to unlock access to ALL audio articles.
A well-trained nose can sniff out the smallest details in a scent. Be it caramel notes in an aged whisky or a citrus edge in fine wine, smells that are otherwise imperceptible can be picked up with experience and training. A new study examining the nose of a species less focused on luxury – the fruit fly – may explain this phenomenon.
Sniffing out scent in the brain
Fruit fly brains, containing just 125,000 neurons, have long been an accessible target for neuroscientists. In 2013, one such analysis found that given the same aroma, certain fruit fly neurons reacted uniformly, while others showed varied responses.
The findings were explained away by the field as a result of interfering noise in the experimental setup. CSHL Associate Professor Saket Navlakha and Salk Institute researcher Shyam Srinivasan revisited these findings. “There were two things we were interested in,” Navlakha said in a press release. “Where is this variability coming from? And is it good for anything?”
The pair and their team decided to investigate further by creating a fruit fly smell model. This showed that the variability unexpectedly originated in a circuit deep in the flies’ brains, which indicated that it was more significant than previous analysis had reckoned.
What makes a reliable cell?
The team next lasered in on the cells that responded consistently to similar smells. They dubbed these “reliable cells”. These represented a relatively small proportion of the overall cell population but were key to helping the flies discern between different smells. Another, much larger group of neurons – those that had varied responses to similar odors – were found to help the flies separate out minute smell differences. The team called these “unreliable cells”.
“The model we developed shows these unreliable cells are useful,” Srinivasan said. “But it requires many learning bouts to take advantage of them.” These cells may be the equivalent of those that help a human sommelier distinguish notes in wine.
The team says that similar cell arrangements may underlie our other senses and how we make decisions based on inputs from those senses. The findings may also help AI researchers construct more accurate algorithms – current models rely only on “reliable” cells, responding only one way to similar inputs. “Maybe you don’t want a machine-learning model to represent the same input the same way every time,” Navlakha concluded. “In more continual learning systems, variability could be useful.”
Reference: doi: 10.1371/journal.pbio.3002206
This article is a rework of a press release issued by Cold Spring Harbor Laboratory. Material has been edited for length and content.