Source neuron identity gave no clear added gain beyond session data for initial 50 Hz change in 249 mouse V1 inhibitory connections

We tested whether another measured output of one neuron helps predict its initial response change during a 50 Hz spike train. We used Allen Synaptic Physiology r2.1 and two mouse V1 projects. The main cohort had 620 inhibitory connections. Of these, 249 connections from 79 inferred donor groups had another eligible output from the same source. Five outer donor folds tested a ridge baseline and residual prediction rules. Four inner donor folds selected all penalties and variance terms. Other measured connections in each test session were explicit inputs, and the target outcome had zero weight in its own prediction. Session measurements reduced mean squared error from 0.042211 to 0.036868, a 12.66% reduction. Its conditional 95% donor resample range was 3.45% to 20.54%. Adding source identity reduced error by 0.55% beyond the session rule, with a range of -4.63% to +5.74%. A second split, stricter train counts, and a stronger cell physiology baseline also gave uncertain added gains. Late train change gave a 0.83% error increase. This is a measurement completion test in sparse slice recording graphs. It does not establish a shared synaptic mechanism or predict dynamics from anatomy alone. Fixed-prediction ranges exclude model refitting uncertainty. The package preserves original code, source hashes, the internal plan and amendments, all model fits, computed predictions, figures, checks, and repeat instructions. A full replay matched all 14 model result files.

short-term plasticity · synaptic dynamics · local recording graphs · mouse visual cortex · source neuron identity · measurement completion · structure and function · donor validation