A fixed rule gave no clear decay prediction gain from other inputs to the same target in 152 mouse V1 connections

We tested whether measured inputs help predict voltage decay at another input to the same cortical cell. We used Allen Synaptic Physiology r2.1. The primary set had 152 connections, 69 target cells, and 47 inferred mice. Five folds kept test mice out of reference fitting. The reference used current kinetics, passive cell properties, cell labels, and distance. The fixed rule added half the mean reference residual from other measured inputs of the same sign to the same target. Mean squared log error fell by 21.7%. The 95% mouse resampling range was -0.6% to +42.7% when the reference was fitted again. Ranges assume independent and exchangeable mice under the selected design. On 77 connections with separate controls in the same recording, target calibration increased error by 4.8%. Fit-quality checks and a source comparison did not establish a clear gain. Quarter correction gave a positive secondary range, but was not the primary rule. The results do not establish a general benefit from target calibration. They concern fitted somatic response decay in selected acute-slice connections. The package has code, source versions and hashes, selections, fitted models, reference predictions, group scores, bootstrap draws, figures, and repeat instructions. Original source measurements remain external.

synaptic kinetics · voltage decay · cortical connectivity · sparse calibration · paired recordings · reproducible benchmark