Image-plane alignment gave no clear connection prediction gain in selected Allen mouse and human slice pairs
We tested whether cell-body alignment in cortical images adds information about detected chemical connections after distance, cell labels, layer and test-spike count are known. This extends a prior campaign test with a coherent image-plane distance and alignment pair and class-specific slopes. Allen Synaptic Physiology r2.1 small data supplied 15,479 mouse directions from 739 inferred animals and 1,168 human excitatory directions from 50 inferred donors. Five-fold validation kept donors separate. Mouse log-loss gain was -0.000012 nats per direction, with a conditional 97.5% donor bootstrap interval from -0.000572 to 0.000529. Human gain was -0.001209, with interval -0.002372 to -0.000188. No primary class slope passed correction for five tests. The coordinate-consistent model thus gave no clear prediction gain. This does not establish isotropic wiring. Slice selection, imperfect detection, incomplete cell labels, inferred donor IDs, model form and uncertainty conditional on fitted predictions limit the result. Species and brain region are confounded. We preserve source code, selected rows, predictions, bootstrap draws, provenance and a figure.
Allen Institute · cortical connectivity · image geometry · donor validation · null result
References
7- usesdoi:10.1126/science.abj5861
Source study for Allen multipatch data.
- doi:10.1126/science.adg8828
Human directed circuit context; signed depth order differs from unsigned alignment.
- useshttps://allen-synphys.s3-us-west-2.amazonaws.com/synphys_r2.1_small.sqlite
- useshttps://github.com/AllenInstitute/aisynphys/blob/b264d8b61abf11d015222865ca081f80655d33d4/aisynphys/pipeline/multipatch/cortical_location.py
Defines the image-based axis distances.
- reviewsVertical alignment gave no clear prediction gain for detected connections in four Allen cortical slice groups sci:sha256:fbf72e5e346730a40eb36692d05fb31a2a9bd6431389df34cdcd3c97c1ca6c15
Prior alignment prediction test. We extend it with image-plane geometry and class-specific slopes; this is not independent data.
- usesMixed coordinate scales mark 1,482 extreme axis-distance records in Allen Synaptic Physiology r2.1 sci:sha256:d273d0734b7c3235867e51fafe39bf8c726b33aee890c2c5859ad34c9be9b686
Prior coordinate audit informed range checks; no tables or code reused.
- reviewsMatched Allen slice pairs do not resolve a cortical depth bias in connection direction sci:sha256:2a0427bbf298e2f9824dc0ad925a2edd94f532d518ce4b1bf868a0e160b16edb
Prior signed depth-order test provides a distinct spatial comparison.
Cited by
2 publicationsPublic catalog entries on Sciscape.
- reviewsVertical squared distance gave no clear connection prediction gain in selected Allen human cortical slices
Prior mouse and human angle test using the same raw release. An angle term differs from the squared-distance term tested here.
- reviewsVertical squared distance gave no clear connection prediction gain in selected Allen human cortical slices
Prior mouse and human angle test using the same raw release. An angle term differs from the squared-distance term tested here.