At 30 Hz, source layer gave no clear added voltage prediction gain after first strength in 991 mouse V1 connections

We tested whether source layer improves second voltage predictions after the first mean response is known. The data had 991 detected excitatory connections, 69 targets, and 39 recording-date groups from developing mouse V1. Stimulation was at 30 Hz. Each test date was absent from model fitting and parameter selection. The baseline used first response, target type, and target layer. The expanded rule added source-layer offsets and strength slopes. Both used nested ridge selection. Mean squared voltage error fell from 0.215636 to 0.214423 mV squared, a 0.56% decrease. In 499 complete refit date resamples, the central range was -19.87% to +10.08%. On the natural-log scale, error rose by 2.38%, with a complete refit gain range of -6.26% to +6.29%. Removing the date with conflicting age labels changed the voltage gain to 4.40%. Its complete refit range was -23.11% to +10.84%. Threshold, weighting, and geometry checks are preserved. The results give no clear added source-layer gain for the tested calibrated rules. They do not establish an absent biological layer effect. The source maps are selected slice recordings, not full anatomical connectomes. The same stimulation repeats supplied both mean responses. Trial noise and a complete animal registry were unavailable. The package preserves source data, code, row audits, every prediction, model states, resample draws, figures, and repeat instructions.

connectome · structure and function · visual cortex · source layer · short term plasticity · voltage prediction · excluded recording dates