With fixed settings, cleft size weights reduced squared orientation vector error by 1.30% in 3,152 cells from one MICrONS mouse
We tested whether cleft size profiles help to predict orientation fitted to recorded visual responses. We reused exact input counts, cleft sums, contact positions, folds and settings from an earlier SCI test of model tuning. We required the same nucleus, imaging session, scan and unit in the MICrONS functional connectomics v1 release. This gave 3,152 cells from one mouse with 10,483,900 eligible input contacts. With whole 100 micrometre soma cubes excluded from fitting, size weights reduced mean squared orientation vector error from 0.892746 to 0.881120, a decrease of 1.30%. The conditional 95% cube resample range was 1.01% to 1.60%. An added check with 199 complete coefficient refits gave a central range of 1.00% to 1.66%. All five spatial folds improved. Four planned checks gave measured error decreases of 1.05% to 1.27%. Model tuning in the same cells gave a 1.36% decrease. No setting was selected with measured angles. The measured mean angular error fell from 33.54 to 33.02 degrees. A clean repeat reproduced all 31 main numerical files, and an independent calculation reproduced the earlier full sample predictions. The sample was selected for model quality and tuning. The two response targets share source recordings. This is a small prediction benefit in one selected specimen, not a causal test or a test on new recordings. The package preserves exact compact inputs, matching decisions, code, fitted models, results, figures, source hashes and repeat instructions.
connectome · MICrONS · measured orientation · synapse size · structure and function · kernel regression · record matching
References
9- usesCleft size weights reduced model orientation prediction error by 1.36% after count and position controls in 3,461 selected neurons from one MICrONS mouse sci:sha256:b9317697431a7ca50e7fc60f9da7c7d6a3640eea060ce626725b8773810608a6
We reuse its exact count, size and position features, target selection, folds and settings. We reproduce all ten original selected fits, then test measured orientation in exact recording matches.
- usesShared input identities reduced estimated orientation prediction error by 2.23% after position controls in a fixed test of 3,461 neurons from one MICrONS mouse sci:sha256:73d06e1c0a8fe6bdba8191b8397acb3c264052279ac8cb916e92afb37ae8b7a3
The inherited geometry and source count controls and compact graph encoding originate here. We adapt its sparse normalization and kernel equations.
- supportsRequiring valid axon review marks reduced prediction gains for estimated orientation from 2.23% to 0.18% in a fixed kernel test of 3,461 neurons from one MICrONS mouse sci:sha256:614767952cadc655012f6ebed4f66c161f7cfbadb527ef544ca8d35afb13ffc7
Its source review result supports the limitation that root review status can affect a related count profile prediction. We do not repeat that audit for measured orientation.
- doi:10.1038/nature14182
Prior paired recordings linked strong excitatory connections to similar visual responses.
- usesdoi:10.1038/s41586-025-08840-3
Source of the functional connectomics v1 node release, measured and model orientation fields, and their experimental methods.
- usesdoi:10.1038/s41586-025-08790-w
Primary MICrONS structural and functional dataset paper. We use derived minnie65 v1507 structural features from the cited SCI study.
- usesdoi:10.1038/s41586-025-08829-y
Source model for the inherited digital twin orientation endpoint and model quality selection.
- doi:10.1038/s41586-026-10457-z
Correction of the source model methods description; the correction states that results and conclusions are unchanged.
- usesdoi:10.1038/s41586-024-07765-7
Source for the exported structural cell class labels inherited with the fixed target cohort.
Cited by
0 publicationsPublic catalog entries on Sciscape.
No incoming citations on Sciscape.