Direct axonal direction gave no clear gain over neuron count in a fixed ridge test of Granger direction in 1,242 mouse region pairs

We tested whether the direction with more axonal projection had the larger spectral Granger score within a pair of mouse brain regions. We joined an IBL source-table export to fitted projection weights from Oh et al. The main set had 1,242 region pairs, 2,452 session comparisons, and 359 sessions. All 213 anatomical nodes and 21,654 positive nonself edges could support indirect routes. Direct anatomical and Granger direction contrasts had rank correlation -0.0426. The recorded neuron count contrast had rank correlation 0.2450. Five pair folds and training-only ridge selection gave count-baseline error 0.113684. Adding direct anatomy gave 0.113738. The error-reduction range from random region weights was -0.948 to 0.848 percent. A baseline with region effects gave a direct-anatomy gain of 0.0233 percent, with a range of -0.103 to 0.148 percent. Walk scores gave small gains with region-weight ranges across zero. Four other anatomical matrix variants extended the test to as many as 1,852 pairs; none gave a direct-anatomy gain. In an added check of 441 repeated pairs, the within-pair count slope was 0.2268, with a region-and-session weight range of 0.0543 to 0.4092. This association does not show a sampling cause. Projection magnitude still had a weak positive relation with Granger magnitude. Different mice, pooled hemispheres, regional averages, and modest repeat agreement limit the test. The result does not exclude structure-dependent function at other scales or frequencies. The package preserves code, source hashes, plans, calculations, predictions, figures, and repeat instructions.

mouse connectome · Granger direction · axon tracing · neuron sampling · structure and function · IBL brain-wide map · directed communication · reproducible benchmark