Connectome Graph Explorer

How to read this

• Base is the Shiu et al. 2024 LIF on the connectome (MaleCNS: brain only, monoamines removed, weights × 0.581, rate APL). On both connectomes, smells keep about 3,000–8,000 neurons firing after the stimulus stops.
• Calibrated replaces the chemical cholinergic LN→PN synapses with eLN–PN gap junctions, because Yaksi & Wilson (2010) showed that this connection is electrical (it survives nicotinic blockers and is lost in shakB²). The persistent state goes away, and odors activate only their own glomeruli's PNs.
• Shuffled keeps each neuron's number of inputs and outputs and the sign of its outputs, and randomizes who connects to whom. If the readout works just as well on it, decoding success says little about the real wiring.
• Limits. Electron microscopy sees chemical synapses only, so gap junctions are placed where eLNs and PNs touch, which is an assumption. At this coupling strength, the excitatory LNs are shunted almost silent, which real eLNs are not. Wind still leaves a tail of about 390 neurons outside the antennal lobe on MaleCNS. On the base model, 0.5–1.0 s decoding uses the persistent state; on the calibrated model, it uses the offset transient, which still carries word identity. The model has no adaptation or short-term depression.
• Each graph shows at most 1,200 of the neurons that fired (the most active ones), each neuron's strongest input, and the strongest remaining edges up to 4,000.