CCM Researchers Win Competition to Map Neural Connections in Fruit Fly Brain

50 largest neurons of the fly brain connectome. Tyler Sloan and Amy Sterling for FlyWire, Princeton University, (Dorkenwald et al., 2024)

The Princeton Neuroscience Institute announced today that Flatiron Institute Center for Computational Mathematics (CCM) researchers Daniel Lee and Lawrence Saul are the winners of FlyWire’s Ventral Nerve Cord (VNC) Matching Challenge. The challenge asked participants to construct a method for aligning the neural networks of a male and female fruit fly. Lee and Saul accepted the award and presented their solution at the Princeton Neuroscience Institute on March 7.

FlyWire is a global effort to use human-AI collaboration to reconstruct the full fruit fly (Drosophila) connectome, a comprehensive map of neural pathways and connections in the fly’s brain. Fruit flies are commonly used to study the brain, and having a more detailed map of their neurobiology — and how it differs in males and females — will help to accelerate future research.

The organizers of the challenge represented the connectomes of male and female fruit flies as large graphs. They then asked the participants to find the optimal way of aligning the nodes of these graphs (i.e., determining which neurons in the female brain correspond to which neurons in the male brain). This is a difficult task because the number of possibilities is vast, and finding the best permutation isn’t straightforward, Lee says.

“Lawrence and I have actually known each other since college, and we’ve worked on projects aimed at optimizing matrices like these before,” says Lee, who is both a visiting scholar in CCM and a professor of electrical and computer engineering at Cornell Tech. “We realized this challenge was a good way to combine our skills.”

“We called our team ‘Old School’ because we thought certain methods from our past research pointed to an elegant solution,” says Saul, the group leader for machine learning in CCM. “In addition, we used MATLAB, which is an older programming language, and relied more on creative thinking than raw computing power.”

The team also saw the competition as an opportunity to showcase the style of research that is done within CCM and the Flatiron Institute more broadly.

“The competition aligned perfectly with the mission of Flatiron in that it distilled an open question in neuroscience into a problem of computational mathematics,” says Saul. “Our methods are available to other interested researchers, and we hope they will be helpful for future studies.”