A membrane time matching term did not reduce electrical call error in a fixed test of 577 mouse V1 Pvalb pairs

We tested whether similar membrane time constants help predict electrical connection calls within a broad cortical cell label. We used Allen Synaptic Physiology r2.1 and selected 577 mouse V1 Pvalb pairs with 66 positive calls from 240 inferred mice. Both reciprocal manual calls had to agree. Both directions needed finite automated pulse coupling estimates. A fixed logistic model controlled individual time ranges, resistance, distance, layer labels, and recording conditions. Five folds kept whole mice apart. Adding the absolute log2 time difference increased test log loss by 2.09%. The 95% fixed-prediction mouse bootstrap range for the increase was 0.60% to 3.87%. A second split gave a smaller 0.37% increase, with a range that included zero. Neither project alone nor five other planned checks showed a clear gain. A broader same-Cre test used 1,715 Pvalb, Sst, or Vip pairs and also gave no clear gain. The full-data regularized matching odds ratio was 0.750, with a 95% range of 0.353 to 1.345 from 1,000 mouse bootstrap refits. That range permits a substantial association. A large imposed matching effect gave a 10.47% test gain. Independent calculations reproduced all 11,990 probabilities, and a clean repeat reproduced the numerical results. These data do not justify this matching term as a useful default in the tested rule. Selection, detection error, a broad cell label, and a small positive sample limit the result. This study does not establish biological independence or measure synchronization. The package preserves code, exact selections, source hashes, fitted states, predictions, resamples, figures, and repeat instructions.

electrical synapse · membrane time constant · Pvalb · mouse visual cortex · structure and function · out of sample prediction · synaptic physiology