With bounded spline models, mean cleft size gave no clear added prediction gain for movie response similarity in two mouse V1 samples

We tested whether mean synaptic cleft size adds information about measured movie response similarity. We used 2,840 V1 connections from one MICrONS mouse and 1,441 from one V1 Deep Dive mouse. Controls included synapse count, anatomical opportunity, cell position, cell class, imaging session, and response reliability. No test neuron occurred in model fitting. An adjusted linear size slope per doubling was +0.00963 Fisher z in MICrONS, with an approximate dyadic 95% interval of +0.00463 to +0.01462. The V1 Deep Dive slope was +0.00220, with interval -0.00316 to +0.00757. The planned MICrONS prediction baseline failed against a training-mean reference. We preserved that result and made stated repairs. With bounded spline functions and expanded ridge tuning, the first-split size gain was +0.548% in MICrONS and -0.057% in V1 Deep Dive. Conditional 95% neuron-weight ranges were -0.097% to +1.215% and -0.490% to +0.406%. All five repaired split ranges included zero in both samples. The evidence gives no clear added size gain for these final models. It does not establish no functional role for size. Two animals, selected cells, different movies, structural proxies, and ranges that omit model refitting limit the result. The package contains code, compact inputs, provenance, all predictions, planned and repaired results, figures, and repeat instructions.

connectome · structure and function · synaptic cleft size · MICrONS · V1 Deep Dive · visual cortex · prediction · neuron exclusion