neuronmimic-data / README.md
cyd0806's picture
Publish full FlyWire and MaleCNS KC skeletons, physical synapses and group indexes
b4143c8 verified
|
Raw History Blame Contribute Delete
6.39 kB
---
language:
- en
- zh
tags:
- neuroscience
- connectomics
- reproducibility
---
# NeuronMimic reusable data (updated 2026-11-01)
This dataset contains full KC anatomy handoffs, earlier compact source-bound tables,
and a deterministic passive cable fixture for reproducing NeuronMimic baselines. It deliberately does
not contain restricted raw MaleCNS/FlyWire connectome downloads or any claim of
biological validation.
The current structure/dynamics requirement matrix is maintained in the parent
BioFly project at `artifacts/wu_20261030_science_gate/WU_REQUIREMENTS_STATUS_20261030.md`.
## Full KC structure package
[`structure_handoff_v1/`](structure_handoff_v1/README.md) adds complete KC-scope
skeleton/contact data: FlyWire5,177 cells/1,881,678 incident synapses and MaleCNS
v1.0 4,064 cells/4,027,967 incident synapses. Both pre/post coordinates, source
annotations and versioned group memberships are retained. FlyWire also includes
contact-to-branch mappings and whole-brain soma/region context. MaleCNS supplies
original v1.0 coarse SWC, detector confidence and site-level transmitter predictions.
This does not include whole-brain EM volumes or non-KC partner skeletons.
Query/export code is an addition to the existing
[NeuronMimic repository](https://github.com/ydchen0806/neuronmimic/blob/master/docs/STRUCTURE_HANDOFF.md).
Read COORDINATES.json before overlaying any points. Groups remain structural
candidates or official type groups, not confirmed biological functional units.
## Latest: P1D-v3 / bilateral geometry
- `p1d_v3/`: original focused left-ap1 experiment, regional-tree tables, frozen
18-cell historical identities, weak-edge controls, failed matched routing,
MaleCNS/BANC source QC and all negative results. Full rerun requires original
BioFly raw inputs; the source-mode scripts are archived on GitHub.
- `p1d_v3_bilateral/`: left148/right133 complete ap1 cohorts, all directed pairs,
exact soma-block assignments, design matrices, model weights, predictions,
seed/fold contrasts, source hashes and figures. Supports a tested standalone
model refit/evaluation from these inputs. Raw geometry extraction is not rerun.
- [Replay instructions and scientific scope](https://github.com/ydchen0806/neuronmimic/blob/master/docs/P1D_REPRODUCE.md).
- [Source code snapshot](https://github.com/ydchen0806/neuronmimic/tree/master/research/p1d_source_snapshot).
Regional tree proximity improved predictions in all10 soma-block folds and both
cross-side directions. This is retrospective evidence within one FlyWire animal,
not external replication or functional validation. Stable8 remains8/965; no new
functional unit is validated. The 18-cell candidate's matched controls failed;
its nested label increment interval includes zero. Every unfavorable result is retained.
Later MaleCNS QC found all31,355 annotation-missing endpoints in the same-release
original neurotransmitter table. The annotation gap persists; a version mismatch
was not established. Historical P0/P1 receipts are unchanged. BANC root_888 is the
frozen snapshot key; current root aliases and unknown side labels remain explicit.
`SHA256SUMS.json` covers released files; all FlyWire/root IDs are strings in model
input/output tables. Source-data/model inference evidence labels remain separate
from the synthetic NeuronMimic Teacher results. No new dynamics were run here.
## Contents
- `physics_teacher/`: historical fixture archive. Its old evaluation metrics are
**INVALIDATED**; use `p0_corrected_20261030/fixture_v2/` instead.
- `malecns/`: historical normalized source-bound aggregate tables. Endpoint annotation
joins are **BLOCKED**: 31,355 edge endpoint IDs do not match even the full source id
column. These tables are not a validated complete CNS or simulation-ready network.
- `flywire_kc_atlas/`: compact all-5,177-cell structural profiles and the
source-bound 20-unit entry registry. This is single-connectome structural
evidence, not independent biological replication.
- `provenance/`: protocol and source manifests.
## Links
- Code: https://github.com/ydchen0806/neuronmimic
- Parent analysis repository: https://github.com/ydchen0806/bio_fly_cyber_lateralization
- Dataset: https://huggingface.co/datasets/cyd0806/neuronmimic-data
The GitHub repository documents the code and the exact analysis boundaries;
this dataset provides the compact inputs and receipts. Raw data remain at their
original public repositories and are not mirrored here.
## Status and limits
`FIXTURE_TEST_ONLY` is synthetic/passive-cable testing. `VERIFIED_REAL_DATA`
means the local source table was hashed and schema-checked. `BLOCKED_MISSING_DATA`
means the data required for a real multi-compartment biological calibration are
not present, especially contact coordinates, branch IDs, conductance values,
and held-out physiological recordings.
## Corrected P0 release
Use `p0_corrected_20261030/fixture_v2/` for reproducible Teacher data, four model baselines, and corrected results. Historical metric files are invalidated; see `HISTORICAL_METRICS_INVALIDATED.md`. All cable experiments are FIXTURE_TEST_ONLY. Code and scientific metric tests: https://github.com/ydchen0806/neuronmimic . Known-equation PhysicsOnly is more accurate than the trained Students; no biological or speed-accuracy advantage is established.
## P1-v2.2 geometry and structure release
`p1_v2_fine_types/` contains the 5,177-cell identity registry, CM/SU crosswalk,
all directed pair split assignments, contact-free tree descriptors, six genuine spatial
fusion ablations, full-parent child membership, model predictions, figures and audit receipts.
See its P1_REPORT.md for conclusions and limits. The 18-cell left-ap1 residual subset
is exploratory, not a validated functional unit. Original stable8 remains unchanged.
Reproduction instructions: https://github.com/ydchen0806/neuronmimic/blob/master/docs/P1_REPRODUCE.md .
The full soma distance matrix is exactly regenerated by the script rather than downloaded.
Raw-source audits and geometry extraction are not rerun by the compact replay; their
original hashes and receipts are provided. Do not treat imported audit receipts as a fresh audit.
Upstream data retain their original source terms; this release does not grant a new license
for the original connectomes. No independent physiology or wet-lab calibration data are included.