With test mice excluded from model fitting, reverse excitation gave no clear added strength prediction gain after same-PV calibration in 53 mouse V1 pairs

Can reverse excitation help predict a missing inhibitory response? We tested open mouse V1 PV–pyramidal recordings. We recalculated responses from source traces and reproduced the reported sample counts and strength relation. The main test used 53 reciprocal pairs from 23 PV cells in 16 mice. Each target inhibitory response was hidden. Other outputs of its PV cell supplied calibration. All fitted coefficients came from other mice. Adding reverse excitation decreased squared log prediction error by 14.8%. The 95% mouse resampling range was −61.6% to +52.3%, with all model fits repeated. Planned controls left this added gain uncertain. Removing the weakest inhibitory responses changed the point decrease to −10.7%. A planned second test used all 88 pairs from 29 mice, with soma distance and no peer calibration. Adding reverse excitation decreased error by 30.2%, with a range of +8.1% to +47.8%. An added check on the same 53 pairs gave lower errors with the global models than with the tested peer rules. Thus, reverse strength helped prediction across mice in this dataset. Its added value beyond the tested same-PV calibration rule remains uncertain. The test concerns detected slice responses, not conductance, complete anatomical wiring, or circuit stability. The package preserves code, compact derived data, source hashes, predictions, resampling draws, figures, checks, and repeat steps.

reciprocal connections · synaptic strength · PV interneurons · mouse visual cortex · prediction · multipatch recordings · excitatory inhibitory circuits · reproducible research