Hierarchical Community Structure of the Adult Drosophila Connectome Reveals Conserved Circuit Archetypes

Biological neural networks can be meaningfully partitioned into sub-networks referred to as communities. Community structure is thought to support functional specialization, adaptability, and cost-efficient wiring. While many studies have focused on inter-regional brain networks, little is known about communities in connectomes mapped at the level of individual neurons and synapses. Here, we analyze the whole-brain Drosophila adult connectome using a nested stochastic block model to uncover its hierarchical community structure. We find that most of the roughly 1500 fine-scale communities–the smallest, and best resolved level of the hierarchy–are spatially compact, generally assortative, and aligned with known biological features such as neurotransmitter identity and cell type. Nonetheless, we find evidence of nonassortative communities, spatially co-localized within the optic lobe and vision-processing pathways. Seeking “functional primitives”–small circuits with functionally narrow feature profiles–we use data-driven clustering to group communities into 45 archetypical meta-clusters based on their spatial, functional, and molecular properties, revealing modular building blocks from which larger, functionally diverse communities are composed. This work advances our understanding of how structure and function are organized in the fruit fly brain and highlights the value of statistical network models in interpreting nanoscale connectomes.