In 2019, researchers at the HHMI Janelia Research Campus in Berlin began brainstorming ways to extract more information from insect connectomes during a walk in the woods. Connectomes provide detailed maps of brain cell connections, but don’t reveal how signals are transmitted between neurons. To address this, the team considered using data on neurotransmitter release to predict how neurons communicate with each other. Neurons release different neurotransmitters to convey specific signals, which can be identified through chemical markers at synapses. The researchers were curious if a computer model could differentiate between synapses where different neurotransmitters are released.
Despite initial skepticism, Jan Funke and his colleagues decided to test the idea by assigning the project to Michelle Du, a high school student starting a summer internship at Janelia. Du learned how to train a neural network to recognize images, a valuable skill for a future computer scientist. Using published data, Du trained the model to predict neurotransmitters and achieved over 90 percent accuracy on test data within a few days. This surprising success prompted Funke and his team to carefully review the data and model to confirm the findings. The model reliably predicted neurotransmitters, but the researchers were unsure how the network was making these accurate predictions.
After confirming the accuracy of their model and ruling out confounding factors, the team developed a method to understand how the network was predicting neurotransmitters. By predicting a neurotransmitter from a known image and altering the image to correspond with a different neurotransmitter, the researchers identified key features used by the network to make predictions. Understanding these distinct traits enabled the team to release their method to the neuroscience community in 2020. This groundbreaking approach allowed scientists to predict neurotransmitters in fruit fly connectomes, providing insights into how neurons influence each other within brain circuits.
Michelle Du, now an undergraduate at Duke University, played a vital role in developing this method, which has since been applied to various fruit fly connectomes by Janelia researchers and collaborators. The ability to predict neurotransmitters in connectomes enhances scientists’ understanding of brain circuit functions and facilitates the generation of testable hypotheses in laboratory settings. Despite initial doubts about the viability of their idea, the team’s innovative approach has had a significant impact on the field of neuroscience. Funke credits the success of the project to Du’s exceptional talent and the willingness to explore unconventional ideas.
Overall, the project’s success serves as a testament to the value of experimentation and collaboration in scientific research. By encouraging curiosity and creativity, the team was able to develop a novel method that has significantly advanced understanding of brain circuitry and neurotransmitter communication. Through the dedication and ingenuity of researchers like Michelle Du and the support of experienced scientists like Jan Funke, groundbreaking discoveries can emerge from seemingly unconventional ideas. The journey from a casual conversation during a walk in the woods to a transformative scientific breakthrough highlights the importance of open-mindedness and persistence in pushing the boundaries of knowledge in neuroscience and beyond.
