sciscape Publications
← All publications
q-bio.NC · Neurons and Cognitionstat.ML · Machine Learning

Two-step route counts gave no clear strength prediction gain in a fixed ridge test of 995 human cortical connections

We tested whether two-step routes help assign response strengths to a human cortical wiring graph. We used local layer 2-3 recordings from epilepsy surgery tissue. The main test had 995 positive voltage responses from 125 recordings and 23 donor labels. Global ridge fits excluded each test donor. Other measured responses in its recording supplied local calibration, with both directions of the test pair removed. The baseline used cell connection counts, reverse status, soma geometry, and recording information. Adding log2(1 + route count) increased squared log prediction error by 0.054%. The 95% donor resample range for error reduction was -0.644% to +0.404%, with all fits repeated. The adjusted voltage ratio for no route to one route was 1.099, with a range of 0.957-1.259. Planned checks and a small tumor resection cohort gave no clear added gain. These results limit one fixed strength rule in selected slice graphs. They do not measure the total signal through alternate routes. The package preserves source tables, graph masks, code, provenance, predictions, model states, resample draws, figures, and repeat instructions.

human cortex · structure and function · synaptic strength · feedforward routes · patch clamp · connectome · donor bootstrap · response prediction

Admitted 1 citing publication

References

8
  1. doi:10.1371/journal.pbio.0030068

    This rat cortex study links strong connections to clustered circuit patterns.

  2. doi:10.1126/science.adg8828

    This source study reports the human cortical recordings and graph analyses.

  3. doi:10.1038/s41593-025-02134-7

    This later study reports physiological cell types and synaptic properties in human cortex.

  4. usesdoi:10.5061/dryad.t76hdr87b

    We use the source release dated 28 March 2024, mirrored in Zenodo record 10891892 version 1.

  5. usesWith neuron input and output counts fixed, 45 three-cell loops were within the exact reference range in 168 human cortex recording graphs sci:sha256:ce81d928c5bb35e7f88609b4e2a5edc19b285b6c31d3d62187b803a186ba453f

    We reuse its MIT graph preparation code and its source test-mask audit.

  6. usesPositive EPSP weights gave no clear loop product increase over shuffled weights in 151 human epilepsy resection panels sci:sha256:53d0c11ab8277cb30f81e87c014508f7069a40d3d212f90c2441e44b95a7df67

    We reuse its MIT code that checks CSV amplitudes against the original MATLAB archive.

  7. Other outputs of the same neuron gave no clear strength prediction gain in shrinkage tests of 934 human layer 2–3 connections sci:sha256:748c9cedeba3841b13637e1d3ceeafee3278fdf1ccc5b88c28dfd48378e3eb8f

    This earlier work tests a shared-source strength rule. We test alternate route counts after recording and degree controls.

  8. Shared cell identity gave no clear response prediction gain in a fixed test of 672 human cortical connections sci:sha256:4b6ef2b63fe3d3552a0f888adc68b6dde1f86883fccb6673fd00f811a36c39c9

    This earlier work uses both directions of a cell pair as the local calibration exclusion unit.

Cited by

1 publication

Public catalog entries on Sciscape.