Fixed wiring summaries leave different growth thresholds in equal-weight models of three selected Allen mouse circuits

We tested whether wiring summaries fix a transient-growth threshold in equal-weight signed linear models. We reused published graphs from Allen Synaptic Physiology r2.1: 247 inferred mice in VisP and 17 inferred human donors in TCx. Exact enumeration fixed each cell input and output count, class connection totals, and reciprocal-pair count. The numerical abscissa could change in five mouse graphs. Fixing unsigned three-cell counts left four; fixing reciprocal-pair class counts left three. No human graph retained variation under the primary constraints. In one six-cell example under the last constraints, two allowed wirings had the same largest eigenvalue real part of 1 but numerical abscissae of 1.151388 and 1.244245. At gain 0.85, both linear systems were stable, but only the latter allowed transient norm growth. This is a model counterexample, not measured neuronal amplification. Independent enumeration and trajectory checks passed. Sparse sampled circuits, slice detection, inferred donor labels, assumed weights, and a common time constant limit the result.

Allen Institute · transient amplification · signed connectivity · exact enumeration · circuit summaries · linear model