Key annotations from: Tony Azevedo, @tuthill.bsky.social , @zandawala.bsky.social‬, @hokuba.bsky.social‬, Steffi Hampel, @andrew-seeds.bsky.social, Kathi Eichler
@jefferis.bsky.social, Michael Pankratz, @fleyes.bsky.social, & Marie Suver.

Renderings by @amysterling.bsky.social and Arie Matsliah

Neurotransmitter predictions from Kevin Delgado, @adjavon.bsky.social, & @janfunkey.bsky.social . Key analysis from Mohammed Osman. @sdorkenw.bsky.social & Forrest Collman for CAVE. (cont'd 2)

Thanks and congratulations to others on the phenomenal team including: Arie Matsliah for Codex, its development, and integration of the BANC. Key technical work from @perlman.bsky.social‬ & Zetta AI. Aelysia for specialist proofreading. (cont'd)

An incredible collaboration with Rachel Wilson & her lab, Zaki Ajabi & @jdrugowitsch.bsky.social @harvardmed.bsky.social‬; Mala Murthy, Sebastian Seung, & the FlyWire team @princetonneuro.bsky.social for their huge proofreading effort; and Ryan Maloney & @debivort.bsky.social for the specimen.

Our data support an architecture of distributed, parallelized, and embodied control, reminiscent of “subsumption architectures” from autonomous robotics, where behavior-centric feedback loops are organized s that they can be combined or subsumed to generate complex or resolve competing behaviors.

* Brain regions involved in learning and navigation supervise these circuits.

* Long-range pathways are organized to permit coordination within and across modules to fine-tune, prioritize, resolve conflicts, and link related behaviors in sequences. This may offer structural substrates for behavioral compositionality.

* Long-range ascending and descending neurons can combine local loops into behavioral modules.

bsky.app/profile/mott...

Using the this influence metric, we find:
* Local sensorimotor loops linking matched sensors and effectors are basic building blocks of behavioral control.

Moreover, Zaki Ajabi developed a computationally efficient method for quantifying the “influence” any neuron has on any other neuron in the CNS. We applied this method to estimate the pairwise interactions between all cells in the CNS, amounting to more than 20 billion influence scores.

The dataset includes ~160,000 cells, segmented nuclei and mitochondria, synapses and neurotransmitter predictions, and annotations linking the CNS to peripheral sensory, motor, and visceral systems.

Info: banc.community
Data: codex.flywire.ai?dataset=banc
Viewer: ng.banc.community/view

The BANC is the first connectome that explicitly links the brain to the nerve cord, & through it, to the body. It offers a new “embodied” perspective for connectomes, one that changes how we think about neural networks for control, behavior, & even artificial intelligence.
bsky.app/profile/mott...

Thanks to those that supported the work including: @wellcometrust.bsky.social‬, @klingensteinorg.bsky.social, @simonsfoundation.org, ‪@sloanfoundation.bsky.social, Searle Scholars Program, the Smith Family Foundation, and the Pew Charitable Trusts.

These results offer a new view of how host cues are processed in the mosquito brain and comparative and evolutionary perspectives on neuronal circuits.

Congrats to the authors and collaborators including Jialu Bao and Wesley Alford who lead this effort!

Using a connectome-informed computational model, @briandepasquale.bsky.social showed that recurrent primary sensory connectivity boosts the downstream neurons output under realistic background odor conditions, meaning recurrent synapses make CO₂ detection robust, even in a noisy world.

Compared to recent fruit fly connectomes, this massive increase in recurrent connectivity and ribbon-like synapses was unique to CO₂ circuitry in the mosquito.

At some of these recurrent sensory neuron contacts, we found ribbon-like presynaptic structures, these had previously only been seen in vertebrates. In vertebrate neurons, ribbon synapses are thought to enable sustained or graded transmission. These could further enhance recurrent signaling.