A connectome is a comprehensive map of all neural connections in a nervous system - every neuron, every synapse, the strength and type of each connection. It's the structural blueprint of a brain, the way a netlist is the blueprint of a circuit.
Producing one requires serial section electron microscopy: brain tissue is embedded in resin, sliced into thousands of ultrathin sections (20-40nm), each section imaged at nanometer resolution, then sections computationally reconstructed into a 3D volume. Individual synapses are visible. Every axon can be traced. The process generates petabytes of raw image data for even small tissue samples.
The computational challenge of reconstructing traced neurons from this data is as hard as the imaging itself. Automated segmentation using convolutional networks has made it tractable - but manual proofreading is still required. A full human brain connectome remains decades away even with current technology.
complete
complete 2023
mapped 2021
unmapped
John White, Sydney Brenner. 13 years of manual reconstruction. The gold standard. OpenWorm reproduced the worm's behavior in simulation using only this connectome + muscle model.
FlyWire project (Princeton/Seung lab) completed the full adult fly brain connectome in 2023. 12 million+ AI-traced neuron segments, community proofreading. First complete connectome of an organism with a complex brain.
Larval zebrafish are optically transparent and genetically tractable. Calcium imaging can record the activity of nearly every neuron simultaneously in a living animal while it behaves. Unique window into whole-brain dynamics.
Google + Harvard mapped 1 cubic millimeter of human temporal cortex (H01 dataset, 2021). 57,000 cells, 150M synapses, 1.4 petabytes. That's 1/1,000,000th of the brain. Extrapolating: a full connectome ~1 zettabyte of image data.
The C. elegans connectome isn't just a curiosity - it's a fully characterized circuit that explains the worm's complete behavioral repertoire. Every neuron has been named and its role characterized:
This circuit produces thermotaxis, chemotaxis, foraging, mating, escape responses, and simple learning. All from 302 neurons. Every connection in this diagram has been verified by electron microscopy.
Connectomes aren't random graphs. They exhibit specific topological properties that appear conserved across species and may be fundamental to how nervous systems work:
High clustering coefficient + short average path length. Any two neurons connected by few hops. Efficient information routing with localized processing. Found in C. elegans, Drosophila, human fMRI networks.
High-degree "hub" neurons are more densely interconnected with each other than predicted by chance. These form a rich club - a densely connected core that integrates information from periphery. In humans: default mode network nodes.
Neurons cluster into modules with dense internal and sparse external connections. Modules correspond roughly to functional regions. Hierarchical modularity at multiple scales.
Most neurons have few connections. A small number are extremely highly connected hubs. Power-law distribution. Robust to random failure, vulnerable to targeted hub attack. Mirrors many natural networks.