With unit weights and free source covariance, two linear input models gave equal covariance on an 81-neuron panel in two adult C. elegans maps
Can target activity distinguish events shared at one chemical synapse site from separate contact events when source fluctuations are unknown? We tested direct linear input models with equal event variance and unit contact weights. We retained all positive chemical edges among 180 native brain labels in eight Witvliet maps. The main test used two adults and an earlier 81-neuron activity coverage panel. The source profile matrix had full row rank on all 81 active targets in map 7 and all 80 active targets in map 8. One target had no input in map 8. Exact rank bounds and saved positive semidefinite source covariance examples showed that the two noise models could give equal target covariance. With independent sources, minimum covariance difference norms remained at 77.3 and 78.7 percent of the original difference. With all active targets and free source covariance, 39.7 and 36.4 percent remained. Full panels had 21 and 19 source cancelling readout directions, but none worked in both adults. Keeping only targets with at least ten total input sites restored equal covariance for panels of 141 and 143 targets. These are model results, not fits to measured activity. They show overlap of covariance families, not that every measured covariance is ambiguous. The package preserves complete chemical edge facts, shared-target counts, explicit model examples, code, four figures, exact rank checks, source hashes and repeat instructions.
connectome · C. elegans · polyadic synapses · input covariance · source noise · identifiability · observation coverage · linear model
References
6- usesdoi:10.1038/s41586-021-03778-8
The eight brain maps and metadata supply the complete retained chemical wiring and site grouping.
- usesTarget pairs at shared synapse sites had higher activity scores under 2 micrometre count controls in two adult C. elegans maps sci:sha256:c1fca56b41d5319793529f402e16cdc87b77ba352de6f76b32e0c89f1afb1c15
We reuse its verified neuron coverage table and its 81-neuron panel rule. We do not reuse activity scores.
- usesWith independent site states, equal rates and unit weights, eight 162-cell C. elegans panels gave maximum readout variance ratios of 2.12–2.60 against independent contacts sci:sha256:ada68f26197953899702b9eda9ee62bdd4a7663aea0c0fbaafccf65bf498f0da
We extend its shared site covariance model from restricted source correlation to arbitrary positive semidefinite source covariance.
- doi:10.1002/dneu.20307
- doi:10.1016/j.sysconle.2022.105172
- doi:10.1016/j.cell.2023.07.035
Cited by
4 publicationsPublic catalog entries on Sciscape.
- Three known release settings isolate a shared release covariance term in a linear model with fixed source statistics, gains and recording noise
This study showed that free source covariance can hide a shared site covariance difference. We test an intervention model.
- usesWith normalized source mode variance at most one, a joint contrast separated site and contact covariance in two adult C. elegans input models
We reuse verified graph facts, the 81-label panel and normalization. We add sharp finite source variance bounds and a joint contrast.
- supportsWith independent site events and unit effects, decoders that ignored sharing lost 9.76–10.50% of optimal squared SNR at the median in two adult C. elegans input models
Its free source covariance examples limit the interpretation of fixed source noise assumptions.
- At unit leak and normalized gap gain one, shared site noise gave 2.1–2.3% more total voltage variance than contact noise in two adult 180-neuron C. elegans models
The free source covariance result limits identification of a shared site mechanism from target covariance. Our source checks are restricted model cases.