A unique and innovative research tool, originally considered a “crazy idea,” was conceived during a walk in the woods and subsequently tested by a high school student. The tool is now widely used by scientists around the world to forecast neurotransmitters in fruit fly connectomes.
Many great scientific ideas start with a walk in the woods.
In 2019, while strolling through the Berlin Botanic Garden, HHMI Janelia Research Campus Group Leader Jan Funke and some of his scientific colleagues began discussing a common issue: how to extract more information from insect connectomes.
These wiring diagrams provide researchers with unprecedented insight into the intricate connections between neurons in the brain.
The article provides detailed information about brain cells and their connections, but it does not explain how the signal from one neuron impacts other neurons in its network.
The researchers speculated whether they could utilize data from past experiments that identified the neurotransmitters released by certain neurons to anticipate the neurotransmitters released by others in the connectome. Neurons rely on neurotransmitters to communicate, and each type of neurotransmitter is responsible for different signals.
Although the human eye cannot distinguish between synapses on neurons that release different neurotransmitters, researchers are exploring this idea further.computer model. Funke and his team were hesitant but decided to give it a shot. Funke was not very hopeful about the project, but at Janelia, he gave it to Michelle Du, a high school student who was interning in his lab for the summer. The project would give Du the opportunity to learn how to train a neural network to recognize images, which is a valuable skill for a future computer scientist, even if the project didn’t succeed.Interning at Funke’s office, Du demonstrated the ability to train the model on published data and assess its performance on test data. Despite Funke’s initial skepticism, the model proved to be more than 90 percent accurate in predicting certain neurotransmitters.
“I was in disbelief,” Funke admits. “The results were surprisingly high.”
Upon further examination of the data and model, Funke, Du, and their team were convinced that the accuracy was not a fluke: the model could indeed predict neurotransmitters. However, the team remained cautious as they did not fully understand how the network was making these predictions.
“I should have been more cautious,” Funke reflects.”Instead of feeling happy, I was actually worried because we didn’t comprehend what was happening,” Funke says.
After eliminating potential factors that could be influencing their results, the team devised a method to comprehend what the network was perceiving that enabled it to make forecasts.
Initially, they utilized their network to anticipate a neurotransmitter from a recognized image, which it achieved successfully. Then, they instructed a separate network to take that recognized image and modify it slightly to generate an image corresponding to the release of a different neurotransmitter — essentially identifying the minimum characteristics that need to be altered for the model to pre rnrnThe research team identified one neurotransmitter over another and developed a separate method to distinguish these unique traits. The team gained a better understanding of the different features their original network used to make predictions from this information. This led to their confidence in releasing their method to the wider neuroscience community in 2020.
According to Funke, “What most of the neuroscience community has seen from this work is the predictions. They were happy to use it, but for us it was very important to make sure it was actually working.”
Five years later, Du is now an undergraduate at Duke University.The technique she helped to develop has been utilized to forecast neurotransmitters in the connectomes of various parts of the fruit fly brain, as well as the adult fly brain connectome. This information is valuable for scientists as it allows them to comprehend how neurons in a circuit influence each other, enabling them to form hypotheses about the brain circuit’s function that can be subsequently tested in the laboratory. “It all began with a somewhat outlandish idea, something that no one was particularly hopeful about. And what do you do with such an idea? You give it to a high school student.”Funke describes the experience as a valuable learning opportunity. She expresses gratitude for Michelle’s exceptional talent.
