Uniform samples of six labels reduced 40 nm pair clustering calls to 54% in 102 mouse stellate cell profiles with conservative ties

We tested how sparse calcium channel maps retain evidence of spatial clustering. We used 7,744 recorded Cav2.1 labels in 149 granule cell and 102 stellate cell membrane profiles from the Rebola et al. dataset, version 3. Each class had four source folders. We sampled six labels uniformly without replacement from every profile. A 40 nm close-pair score used each profile's own polygon as its uniform reference. Its equal-folder mean stayed near 1.46 in granule cells and 23.1 in stellate cells. Yet one-sided tests at p <= 0.05 with conservative ties called clustering in 100% of full stellate maps and 54.5% of six-label samples. The source-folder range was 48.7% to 61.2%. Random ranks within ties raised the six-label value to 71.7%. A nearest neighbor test retained 83.1%. Added Gaussian position error of 20 nm per axis reduced six-label pair calls to 35.9%. Removal of labels outside outlines gave 54.7% without added error. An exact finite sampling formula gave median minimum budgets of 16 granule-cell labels and 14 stellate-cell labels for relative SD at most 25% in the recorded pair fraction. The original study linked these classes to different release properties through separate measurements and models. Our results concern sampling of recorded molecular maps, not release at individual profiles. The classes have different outline units and scales. The contribution is an open coordinate benchmark, a checked precision calculator, and a controlled test of two spatial readouts. Source bytes, provenance, code, results, figures, and offline repeat instructions are preserved.

calcium channels · synaptic release · spatial point patterns · sparse labeling · finite sampling · cerebellum · immunogold · reproducibility