In 147 ganglion cells from one mouse retina, depth profile similarity gave no clear added gain for movie response similarity in a ridge test with chirp and quality controls

We tested whether similar dendrite depth profiles predict similar natural-movie responses after a uniform-light response is known. We used matched electron microscopy and calcium data from Eyewire II. The main panel had 147 ganglion cells and 2,206 cell pairs in five fields of one mouse retina. We compared responses from different repeats of the same five movie clips. This limits shared noise within a repeat and avoids division by an estimated signal variance. Each outer test excluded one complete imaging field. Inner field tests selected the ridge penalty. The baseline used soma distance, chirp response similarity, and measured response quality. Adding depth profile cosine similarity reduced mean squared test error from 0.0165525 to 0.0165403, a 0.074% decrease. Resampling the five fixed field scores gave a central 95% range from a 0.288% increase to a 0.330% decrease. This range does not measure variation among mice. Three fields improved and two worsened. Eighteen of 199 cell-label controls gave at least as much gain. Using correlation of mean responses gave a larger 0.405% decrease. A source check also placed the 80 depth bins from -15 to 25 micrometres, with 0.5 micrometre width. These results limit this scalar depth feature and this conditional prediction test. They do not test all structural features or direct synaptic weights. The package has independent code, source versions and hashes, aggregate results, figures, numerical checks, and repeat steps. Source data remain external under their original terms.

retina · structure and function · dendrite depth · natural movies · calcium imaging · Eyewire II · response reliability · connectomics