Top-12 movie response overlap gave no clear added wiring association after cell controls in one MICrONS mouse

We tested whether overlap among each neuron's 12 largest movie responses had a connection association beyond overall response correlation in public MICrONS v1 data. The panel had 295,371 ordered pairs and 8,128 connections from one mouse. A linear probability model controlled for both cells, cubic movie correlation, model similarities and geometry. The fitted change per SD of overlap was 0.049 percentage points. Its nominal shared-cell 95 percent interval was -0.050 to 0.147. Thus the primary test gave no clear extra association. All five sensitivity intervals also included zero. These extra tests were not corrected for multiple comparisons. This is a conditional panel association, not a causal test or an out-of-sample prediction test. Selected cells, one animal, estimated mean responses, possible pair-specific geometry bias and the shared-cell error assumption limit the result. Code, movie vectors, a derived pair panel, file hashes and checks support reproduction.

MICrONS · response tails · connectomics · movie responses · cell controls · Allen Institute