The trade-off between fractal morphology and function of Drosophila neurons

Studying correlation between structure and function in brain networks has long been a central and compelling topic in neuroscience. Using the FlyWire project's fully digitized Drosophila melanogaster brain map, we analyzed a total number of 125330 3D skeletons of neurons from various functional super-classes and classes. By separating axonal and dendritic arbors of individual neurons we compare several morphological and topological properties, including cable length, fractal dimension, number of nodes, maximum width, and tree height. The fractal dimensions of neurons can reach up to 2.4, with dendritic arbors generally more complex than axonal ones, except in the afferent flow, where the situation is reversed. Correlating fractal dimension with cable length reveals that neurons from the same classes cluster in state space. These clusters densify and separate more clearly as we go deeper down the hierarchical annotation schema, indicating that the brain specializes fractal morphology for distinct functional classes.