The FlyWire connectome: neuronal wiring diagram of a complete fly brain

The team worked from high-resolution data acquired by electron microscopy. Electron microscopy is still the only tool available for a complete understanding of connectivity at synapse resolution but comes with the notable disadvantage that the neurons are not labelled, leaving an unsegmented, undifferentiated web. Untangling this web requires a multi-step process of segmenting the data, which has been greatly aided by computer and AI automation, proofreading to catch any mistakes, reconstructing the elaborate processes of the neurons (see example neurons on this page), assigning connections between them, and annotating the dataset to fit the neurons into recognized cell classes.

Today, the group announces the complete connectome of an adult female fly brain with carefully curated annotation of the neurons and over 8,000 cell types. This is accompanied by a statistical analysis of the structure of the connectome.

Particular attention has been given to the visual system because most of the fly brain is dedicated to vision. This is the first time the cell types and connections of a biological visual system have been revealed in their entirety.

In total, the group has identified around 140,000 neurons — 98% of which have been typed — and over 50 million synapses.

Now, with connectome in hand, the goal of understanding how the fly nervous system generates behaviours becomes much more tractable. The package of papers contains examples of how the resource is already being used to guide experiments.

One strategy is to use the connectivity data alone to predict phenomena not yet discovered. A good example is the paper from Seung that uses connectivity data to identify a new neural circuit, and predict its role in visual function.

A different approach is to use the resource to build a full network of connected cells from experimental observations, which are typically conducted only at key nodes in the circuit. Sapkal et al. use this method to describe a neural circuit for halting behaviour.

In addition to aiding in experimentation, the resource facilitates new computational models of the fly brain specifically and theories of how nervous systems work generally. Here, the package of papers gives examples of two extremes.

Shiu et al. take the entire connectome and transform it into a model spiking network that uses reasonable assumptions about how neurons integrate inputs, bringing us closer to what might be called a ‘digital twin’ of the fly brain. This approach can be used to mimic the effects of perturbations to the circuit and make predictions about the ultimate changes to behaviour.

However, the nervous system can be broken down into subcircuits, so Pospisil et al. ask how much depth do we really need to understand the nervous system? They set out to define an ‘effectome’ from the connectome.