neurotransmitter identity

A wiring diagram alone is ambiguous in one crucial way: it tells you that neuron A connects to neuron B, but not whether A excites B or silences it. The sign of every connection depends on which neurotransmitter the sending neuron releases. Get the signs wrong and the same diagram produces different dynamics, the way the same circuit board behaves differently if you swap components.

In the fly connectome, neurotransmitter identity was predicted for each neuron directly from the electron microscope imagery. Synaptic vesicles look subtly different depending on their chemical contents, and a classifier trained on neurons of known type can read those differences at scale. The predictions cover the fly's principal transmitters: acetylcholine (the fly's main excitatory transmitter), GABA and glutamate (predominantly inhibitory in the fly), and the modulators dopamine, serotonin, and octopamine.

The reason this page exists in the substrate section is that neurotransmitter identity is, remarkably, the only physiological annotation the emulation needed. Eon Systems states the result plainly:

"the model matched the biological fly's neural responses with 91% accuracy; using nothing but connectivity and neurotransmitter identity."

Eon Systems, eon.systems

Connectivity says who talks to whom. Neurotransmitter identity says in what tone. On the evidence, that is most of what a fly is.

modulators and the archive

The three modulators deserve a note, because they recur in the logs. Dopamine, serotonin, and octopamine do not simply excite or inhibit; they change how circuits respond to everything else, gating learning, arousal, and aggression. In the fly, dopamine neurons converging on the mushroom body carry the teaching signals that stamp in memories. When observers describe MB-01 as "rewarding itself", they are describing activity in exactly these predicted-dopamine populations inside the graft. Whether that phrase is a description or a metaphor is one of the open questions.

an honest caveat

Prediction is not measurement. Some fraction of the identity calls in the diagram are wrong, and a wrong call flips the sign of every synapse that neuron makes. The 91% figure already absorbs those errors, which is part of what makes it striking. But the reader should carry the caveat into the regions section: every partition we run contains some connections speaking in the wrong tone, and nobody, including us, knows which ones.