Connectomics Explained

What a Connectome Actually Contains

A connectome is more than a simple list of which neurons connect to which. A complete connectome maps every neuron in a nervous system, identifies every synapse between them, classifies each synapse by type (chemical or electrical, excitatory or inhibitory), and records the morphology (shape and branching pattern) of each neuron. In practice, connectomes also capture information about the spatial location of neurons and synapses, the size of synaptic contacts (which correlates with synaptic strength), and sometimes the neurotransmitter type used at each synapse.

This level of detail is critical because the function of a neural circuit depends not just on which neurons are connected but on how strong those connections are, whether they excite or inhibit their targets, where on the target neuron they make contact (soma versus dendrite), and how they are arranged spatially. Two circuits with identical connectivity graphs but different synapse types or locations can produce completely different computations.

The distinction between a connectome and other brain maps is resolution. Diffusion MRI tractography maps the large fiber tracts connecting brain regions but cannot resolve individual synapses. Functional connectivity, measured by fMRI, reveals which brain regions tend to be active together but says nothing about the underlying anatomical connections. Only electron microscopy-based connectomics provides the synaptic-resolution wiring diagram that brain simulation requires.

How Connectomes Are Built

Building a connectome is an extraordinary technical challenge. The process begins with tissue preparation: a brain or brain region is chemically fixed and stained with heavy metals that make cell membranes visible under an electron microscope. The tissue is then cut into ultrathin sections, typically 30 to 50 nanometers thick, using a diamond knife or focused ion beam.

Each section is imaged by scanning or transmission electron microscopy at nanometer resolution. A cubic millimeter of brain tissue, roughly the size of a grain of sand, produces on the order of one petabyte (one million gigabytes) of image data. The entire fruit fly brain required imaging roughly 7,000 sections, generating datasets measured in hundreds of terabytes.

The images must then be aligned across sections and segmented to identify individual neurons and synapses. Early connectomics projects relied on manual tracing by human annotators, an approach that was painstakingly slow. The C. elegans connectome, comprising just 302 neurons, required over a decade of manual work by Sydney Brenner and his colleagues, earning Brenner a Nobel Prize in 2002.

Modern connectomics has been transformed by machine learning. Deep neural networks trained on human-annotated examples can now segment neurons in electron microscopy images with accuracy approaching that of expert human tracers. Google collaboration with the Lichtman Lab at Harvard developed flood-filling networks that propagate segmentation through three-dimensional image volumes, dramatically accelerating the reconstruction process. Human proofreaders still verify and correct the automated results, but the ratio of machine to human effort has shifted enormously.

Landmark Connectomes

C. elegans (302 neurons, ~7,000 synapses): Completed by John White, Sydney Brenner, and colleagues in 1986, this remains the only connectome of a complete adult animal nervous system that has been available for decades. It has been used to model locomotion, chemotaxis, and thermotaxis, and it provided the template for the OpenWorm simulation project. Despite having the complete wiring diagram for nearly 40 years, researchers still discover previously overlooked connections and functional relationships in this tiny nervous system.

Drosophila melanogaster (roughly 140,000 neurons): In 2024, the FlyWire consortium published the complete connectome of an adult female fruit fly brain, the largest complete connectome to date. A team of over 200 scientists combined automated segmentation with community proofreading to map every neuron and synapse. Fully proofread connectomes now exist for both the male central nervous system and the female brain. This dataset has already enabled discoveries about the neural circuits underlying visual processing, navigation, aggression, and courtship behavior.

Mouse cortex (partial): The largest densely reconstructed volume of mammalian brain is a one-cubic-millimeter sample of mouse visual cortex, containing approximately 120,000 neurons and 523 million automatically detected synapses. This "MICrONS" (Machine Intelligence from Cortical Networks) dataset, released by the Allen Institute and collaborators, provides unprecedented detail about the connectivity patterns within a small region of mammalian cortex. However, this represents a tiny fraction of the full mouse brain, which contains roughly 70 million neurons.

Larval zebrafish: The larval zebrafish brain, with roughly 100,000 neurons and the advantage of optical transparency (allowing calcium imaging of the entire brain in a living animal), has emerged as an important model for connectomics. Partial connectomes of the zebrafish hindbrain and spinal cord have revealed circuit motifs underlying swimming behavior.

What Connectomes Reveal

Connectomics data has already produced several fundamental discoveries about brain organization. First, neural circuits are far more structured than random networks but far less regular than engineered circuits. They contain specific connectivity motifs, patterns of connections among small groups of neurons, that appear more frequently than expected by chance. These motifs, such as reciprocal connections between pairs of neurons or convergent input from multiple sources onto a single target, are believed to serve specific computational functions.

Second, the relationship between structure and function is both strong and nuanced. In Drosophila, the connectome has revealed that neurons with similar functions tend to have similar connectivity patterns, but there is also significant variability between individual animals, suggesting that brains may use different wiring solutions to achieve similar behavioral outcomes.

Third, connectomics has revealed the importance of connection types. In the fly brain, the ratio of excitatory to inhibitory synapses, the distribution of synapse sizes, and the patterns of convergence and divergence all follow specific statistical rules that constrain the computational capabilities of neural circuits.

Limitations and Open Questions

Connectomics provides the structural wiring diagram of a brain, but structure alone does not determine function. The same anatomical connection can have different functional effects depending on the activity state of the network, the levels of neuromodulators, and the history of recent activity (short-term synaptic plasticity). A connectome is a static snapshot of a dynamic system, and understanding the full system requires combining connectomics data with functional recordings, pharmacological manipulations, and behavioral measurements.

Scaling connectomics to larger brains remains a formidable challenge. A complete mouse connectome would require processing roughly 100,000 times more image data than the fruit fly connectome. A human connectome at synaptic resolution is currently beyond technical reach, both in terms of the tissue processing and imaging time required and the computational resources needed for reconstruction.

Despite these limitations, connectomics has become an indispensable tool for neuroscience and for the artificial brain research that depends on it. Every brain simulation that aspires to biological realism must be constrained by connectomic data, and the rapid growth of available connectomes is making increasingly realistic simulations possible.