Mouse brain regions farther from the retina in the Allen connectome had less reliable visual spike responses in 39 mice, but distance-preserving rewired connectomes predicted this equally well

Question: a prior SCI study found that the response reliability of mouse brain regions decreases with weighted connectome distance from the retina (D_ret). Does this rule generalize to new mice, stimuli and time scales, and does it need the specific wiring? Data: Allen OpenScope Barcoding (14 new mice; full-field flicker and flickering gratings) and Vision2Hippocampus (25 mice; moving bar), Neuropixels; Knox 2019 mesoscale connectome with Allen retinal tracer entry. Unit reliability is the odd-even repeat correlation after random thinning of spikes to an equal count. The plan was frozen before region values were computed. Results: region reliability decreased with D_ret in all four planned tests (Spearman rho -0.52 flicker, -0.43 gratings, -0.46 fine band below 25 ms, -0.43 bar; 41-44 regions; all mouse-bootstrap samples negative). It stayed after control of distance from LGd, firing rate and mouse, and inside the isocortex (rho -0.82 to -0.87). Region values agreed across data sets (rho 0.88). But the planned wiring-specificity test failed: 1,000 rewired connectomes that keep each target's input shares by source distance gave nearly the same correlations (mean rho -0.40 to -0.42; p 0.19-0.42). Exploratory: inside the isocortex the real wiring did better than the rewired graphs (p 0.002-0.013), and a simple hop count was the best predictor (rho -0.58 to -0.68). Limitations: an average-mouse connectome model; one null design; the result depends on the choice of retinal entry regions; visual stimuli only; exploratory results need independent confirmation.

structure-function · connectome · response reliability · retina · Neuropixels · Allen Institute · OpenScope · mesoscale connectivity · null model · mouse visual system · replication