A shared opponent inhibition slope raised synapse count test loss by 1.0% in one mouse PPC sample

We tested whether a reported functional wiring rule improves synapse count prediction for cells excluded from fitting. We used matched anatomy and task selectivity from one mouse posterior parietal cortex sample. The cohort had 747 candidate pairs, 13 inhibitory cells, and 114 excitatory cells. Each main test excluded every pair that contained one inhibitory cell. A Poisson count model used physical overlap and separate direction intercepts. Adding one common opponent inhibition slope raised test deviance from 488.1303 to 493.1732, a 1.03% increase. Eleven of 13 cells had higher loss. Separate direction slopes gave a 0.87% gain, with a descriptive fixed-prediction cell bootstrap range from -2.24% to +2.78%. Geometry and marginal selectivity controls gave no added gain for the common slope. An exploratory rate analysis found that the inhibitory-to-excitatory opposite/same preference ratio changed from 1.213 among connected pairs to 0.840 among all 550 candidate pairs. An exact overlap accounting explains this sample change. The all-pair ratio range was 0.577 to 1.228. These results limit the use of connection-only correlations as count rules. They do not reject opponent inhibition or show a population reversal. The package preserves plans, code, source hashes, model outputs, all fold losses, resamples, and figures. Independent fits agreed, and a fresh source rebuild reproduced the numerical results exactly.

opponent inhibition · posterior parietal cortex · connectome · structure and function · synapse count prediction · cell exclusion · physical contact opportunity