Seven of 15 testable Allen mouse inhibitory circuits need changes above factor 1.25 to fit a source and target amplitude model

We tested whether resting synaptic response amplitudes equal a source-cell factor times a target-cell factor in Allen Synaptic Physiology r2.1 slice data. The primary mouse inhibitory sample had 1,060 detected connections in 619 recordings. Only 15 recordings had a loop in the source and target measurement graph and could test this model. These gave 25 independent algebraic constraints. Seven of the 15 needed changes above a factor of 1.25 on the measured amplitudes to permit an exact model fit. The largest required factor was 2.095. A linear program gives the smallest factor, and stored primal and dual certificates verify the result. At a 50 microvolt amplitude floor, five of 13 testable recordings still needed a factor above 1.25. These are bounds on changes to fitted values, not measured error limits. Sparse graphs, selected slices, unknown fit errors, and few human constraints prevent a population or mechanism claim. Code, extracted data, provenance, a figure, and independent checks are included.

Allen Institute · synaptic strength · source and target factors · graph constraints · error bounds · circuit models