Data sources
Verified 2026-09-26 by listing buckets, downloading files, computing on them and
querying APIs. Every SHA-256 below is of a file on this machine under
.cache/data/ and its base64 MD5 matched the GCS object hash. Anything not
directly verified is marked UNVERIFIED. Full notes with schemas, snippets and every
column: docs/research/data_malecns.md, docs/research/data_flywire_and_others.md.
0. Summary
| Dataset | Version(s) | Bulk access (no token) | Live API (token) | Neurons | Licence | Our registry status |
|---|---|---|---|---|---|---|
| MaleCNS | v1.0 (2026-06-08; v0.9 2025-10-05) | gs://flyem-male-cns/v1.0/connectome-data/flat-connectome/ Feather |
neuPrint male-cns:v1.0 |
166,700 | CC BY 4.0 | pin now |
| FlyWire / FAFB | v783 (latest, Oct 2023), v630 (June 2023) | gs://flywire-data/codex/data/fafb/{630,783}/ CSV.gz, public (the Codex web page is only a login front for the same bucket); Zenodo 10676866 (v783 Feather) |
CAVE flywire_fafb_public (token; even the info endpoint redirects to login now) |
139,255 / 127,978 | papers + Zenodo CC BY 4.0; Codex bucket has no LICENSE object; FlyWire ToS text UNVERIFIED | pin now (both) |
| Shiu 2024 model inputs | v630 export of 2023-03-23 (+ v783 twin) | GitHub philshiu/Drosophila_brain_model raw files | – | 127,400 | MIT | pin now (sim golden input) |
| hemibrain | v1.2.1 (neuPrint); v1.2 flat export | gs://hemibrain/v1.2/ (public) |
neuPrint hemibrain:v1.2.1 |
21,739 traced non-cropped | CC-BY (version UNVERIFIED on page; paper CC BY 4.0) | pin in M1 |
| MANC | v1.0 (flat export), v1.2.1 (neuPrint, Codex, sjcabs), v1.2.3 | gs://flyem-manc-exports/v1.0/ (public); no v1.2.x flat export except synapse partners in gs://manc-seg-v1p2/; Codex manc/1.2.1/neurons.csv.gz |
neuPrint manc:v1.2.1 |
23,759 / 23,665 traced | CC-BY | pin v1.2.1 via Codex neurons + neuPrint edges (or sjcabs) in M1 |
| BANC | v888 (publication, 2026-04-16), v626 (preprint) | gs://lee-lab_brain-and-nerve-cord-fly-connectome/ (public) and Harvard Dataverse DVN/7WTH1N v3.0 (CC BY 4.0); Codex banc/888/ |
CAVE brain_and_nerve_cord_public (token) |
156,012 proofread+rough (meta); 158,262 (Codex) | CC BY 4.0 | pin in M1 (was "TODO" in the brief; it is fully public) |
| FANC | public exports v1237, v1444; BANC cross-refs v1116 | gs://lee-lab_female-adult-nerve-cord/CAVE/v1444/ parquet (public) |
CAVE fanc_production_mar2021 restricted (PI must email the lab) |
21,978 cell IDs | none stated | adapter only; no cell types in public export |
| sjcabs compiled_data | BANC 888, FAFB 783, MANC 1.2.1, hemibrain 1.2.1, MaleCNS v0.9 | gs://lee-lab_brain-and-nerve-cord-fly-connectome/compiled_data/ (public) |
– | – | code MIT; data "CC-BY by the respective creators" | vocabulary adopted; files not primary |
Two access facts that change the plan relative to the brief: FlyWire bulk data needs no login, and BANC is fully public under CC BY 4.0 with a compiled meta table that already carries cross-dataset type and match-ID columns.
Token status (2026-09-26): neuPrint tokens issued before 2026-08 are invalid
(auth migration); the public neuPrint server answered anonymous Cypher for
male-cns:v1.0, hemibrain:v1.2.1 and manc:*, but neuprint-python refuses to
construct a client without a token and this openness is undocumented. CAVE info
and materialization endpoints for FlyWire, BANC and FANC all redirected to login.
1. MaleCNS v1.0
- Site: https://male-cns.janelia.org (download/, release/). Release notes are two lines: v1.0 (June 8, 2026) "Minor proofreading changes", "Refinement of neuron annotations"; v0.9 (October 5, 2025) initial release. No counts or thresholds on the site.
- Paper: Berg S, Beckett IR, Costa M, Schlegel P, Januszewski M, … Jefferis GSXE (111 authors). Sexual dimorphism in the complete Drosophila male central nervous system connectome. Cell 189(18):5504–5526.e15, 2026. doi:10.1016/j.cell.2026.08.015. PMID 42691995. CC BY 4.0. Preprint bioRxiv 10.1101/2025.10.09.680999. Companion Cell papers: visual pathways 10.1016/j.cell.2026.08.014, gustatory connectome 10.1016/j.cell.2026.08.016. (Cell full text is Cloudflare-blocked from here.)
- Licence: "The Male CNS dataset is licensed under CC-BY" (footer links CC BY 4.0). No further terms.
- neuPrint:
https://neuprint.janelia.org, datasetmale-cns:v1.0(v0.9 still served). Alsoneuprint-cns.janelia.org(401 anonymous). Clio, Cell Type Explorer, Dimorphism Explorer, NeuronBridge. - Bucket:
gs://flyem-male-cns/v1.0/connectome-data/flat-connectome/; HTTPShttps://storage.googleapis.com/flyem-male-cns/.... Neo4j database directory (251.8 GB, Neo4j 4.4.16) atv1.0/database/neo4j/; neuPrint input CSV/Feather atv1.0/database/neuprint-inputs/(194 GB).
| File | Bytes | SHA-256 | Rows | Default? |
|---|---|---|---|---|
body-annotations-male-cns-v1.0-minconf-0.5.feather |
14,483,314 | 2177e246113e4cfbf1e7772ec37c6da1955ff22e8063d0b1f833101f99a9a3b2 |
211,577 × 36 | yes |
body-neurotransmitters-male-cns-v1.0.feather |
43,282,834 | 95c9289220663abeb3409f3ad9e5a7f8a53f8093f5139d15502cd08da8879621 |
1,835,518 × 10 | yes |
connectome-weights-male-cns-v1.0-minconf-0.5.feather |
1,051,241,946 | e35da783d1c686b2b58b3b87cd6a403ae43bfcfba8bff28e08ef752c1a56afc1 |
151,856,684 × 3 | --level weights |
connectome-weights-…-traced-only.feather (undocumented) |
508,025,642 | UNVERIFIED (not downloaded) | 25,563,197 × 5 (type_pre/type_post added) |
candidate default neuron graph |
body-stats-male-cns-v1.0-minconf-0.5.feather |
778,062,826 | ca5dc83a26382ae70c8d8f42fc09ce2dbc1af7c03f3a001a1936b5e142540647 |
88,384,522 × 11 | optional |
tbar-neurotransmitters-male-cns-v1.0.feather |
2,651,680,218 | not downloaded | 45,656,140 (7 nt_*_prob columns, rows sum to 1) |
--level nt-probs |
syn-partners-…-minconf-0.5.feather |
6,777,179,098 | not downloaded | 311,833,243 | --level synapses |
syn-points-…-minconf-0.5.feather |
13,061,489,098 | not downloaded | 357,489,383 | --level synapses |
Schema essentials (annotations): bodyId; status (Traced 165,122 / Orphan / Glia / Unimportant / Assign / Anchor) and ordered statusLabel (31 categories); superclass (27 values; non-null = 166,700 = the paper's neuron count); class, subclass, type (11,751 distinct), instance, group, supertype; somaSide L/R/M, rootSide, somaNeuromere, somaLocation [x,y,z] 8 nm voxels; cross-refs flywireType (143,156 bodies), hemibrainType (32,919), mancType/mancBodyid/mancGroup/mancSerial; itoleeHl, trumanHl, birthtime; dimorphism (4 values), fruDsx (6), synonyms, matchingNotes; entryNerve/exitNerve, receptorType, assignedOlHex1/2, vfbId.
NT file: body, cell_type, total_nt_predictions, predicted_nt, predicted_nt_confidence, ground_truth (85,484 rows), celltype_*, consensus_nt. No per-transmitter probabilities at body level. Traced-body predicted_nt: ACh 94,946, Glu 28,055, GABA 20,218, unclear 14,365, DA 4,443, His 2,026, 5-HT 465, OA 102.
Weights: body_pre, body_post, weight int64; no ROI column; sum 311,833,243 = total PSDs; Traced→Traced 25,563,197 edges / 124,025,046; ≥5: 7,622,864 edges (all segments).
Thresholds: minconf-0.5 = synapse predictions with confidence < 0.5 excluded before export (flyem-snapshot min-confidence); presynapses are already ≥0.70, so the filter bites on postsynaptic confidence only. neuPrint Meta: postHighAccuracyThreshold 0.5, postHPThreshold 0.7. Hence flat weight = neuPrint weight.
ROIs: 144 primary neuPrint ROIs (brain neuropils L/R, VNC neuropils, nerves), 5,619 total incl. optic-lobe column ROIs. The flat weights file has no ROI split; per-ROI weights need syn-partners (primary_post) or neuPrint.
Caveats: 58 % of PSDs sit on unannotated fragments; 1,991 superclass-bearing neurons are "Out of scope"; 516 Traced bodies lack a superclass; type count 11,751 vs paper's 11,710 (UNVERIFIED cause); dimorphism tallies in the file do not reproduce the abstract's counts.
Citation to emit: Berg et al. 2026 (Cell) plus, for NT predictions, the paper's Methods (the classifier lineage is Eckstein et al. 2024; UNVERIFIED whether a separate MaleCNS NT paper exists).
2. FlyWire / FAFB (v630, v783)
- Latest public snapshot v783 (Codex FAQ "v783 – Oct 2023 [latest release]"; bucket has only
630/and783/; flywire_annotations v3.1.0 of 2026-07-21 is still 783-based; Zenodo says 783 is latest). Whether CAVE has newer materializations is UNVERIFIED (token). - Bulk, no login:
https://storage.googleapis.com/flywire-data/codex/data/fafb/{630|783}/<file>(this is the URL template in Codex's own loader). Codex web downloads need Google sign-in +api_tokenbut front the same bucket. Files are refreshed in place without version bumps (e.g. 783classification.csv.gz2026-02-24), so the registry pins MD5/SHA-256 per file and records the GCSupdatedtime. - Zenodo 10.5281/zenodo.10676866 (CC BY 4.0):
proofread_connections_783.feather852 MB (pre/post root, neuropil, syn_count, per-NT avg),flywire_synapses_783.feather9.5 GB (hascleft_score),proofread_root_ids_783.npy. - Annotations:
flyconnectome/flywire_annotations(Schlegel 2024 supplements; v3.1.0 2026-07-21):Supplemental_file1_neuron_annotations.tsv139,248 rows × 31 (cell_type,hemibrain_type,supertype,top_nt,top_nt_conf,known_nt,dimorphism,fru_dsx,vfb_id,fbbt_id, …);Supplemental_file5_hemibrain_meta.csv25,397 hemibrain bodies. - CAVE: datastack
flywire_fafb_public; token fromhttps://global.daf-apis.com/auth/api/v1/create_tokenafter accepting FlyWire terms.
Key files on disk (SHA-256 in the research note for all 40):
| File | Bytes | SHA-256 | Rows |
|---|---|---|---|
630/neurons.csv.gz |
1,572,661 | a0be4670993d86cea0a4724ba24f874c9d867a24f4e248ff30dcf9b6565d81bb |
127,978 |
630/classification.csv.gz |
1,011,550 | b2cf34ef76f5750e647334b56c73ae7f49c96f0fee3371e8a52d26c835a3b499 |
127,979 (one id not in neurons.csv) |
630/connections.csv.gz |
26,124,953 | 68144490b8585fb6b1bf5979d923a3b1081b435cd52b7f82daed1beb6b84781a |
3,794,615 rows; 2,613,129 pairs |
783/neurons.csv.gz |
1,679,884 | 6a6b3759e635f0f35a677d169052362131ec61d95f55919298b55c43fce4e719 |
139,255 |
783/classification.csv.gz |
934,402 | e946b552f4056dfc977707be0674609832c3f64332a22d69dc0d9615e7aae663 |
139,255 |
783/fw_and_hemibrain_types.csv.gz |
1,043,857 | 193f266279f20cfd816e3631a9225cf887d3424d337454a20f09d10916472bd0 |
139,255 |
783/consolidated_cell_types.csv.gz |
901,707 | 8aba246d71dc40361677493629972ce3883048c3d02010adc42bda22962a1a2d |
138,327 |
783/cell_stats.csv.gz |
2,526,548 | bd5879e1b5df964bea2f3ca5316348d4276ce2ccaac283f0e36583c04fbd3d8e |
139,246 (9 neurons lack stats; not a neuron-count source) |
783/connections.csv.gz (= connections_buhmann.csv.gz, identical MD5) |
50,289,304 | d49dd692e59e153aa3c83f5257bfc0eff51247b86d7bb183386c6d1622c70fc9 |
3,869,878 rows; 2,700,513 pairs |
783/connections_princeton.csv.gz |
68,456,801 | 445f996bf6c4b1803b9ba186189138a3061ff8623aa94c0abcf38af30a5bd48b |
5,342,446 rows; 3,732,460 pairs |
783/connections_no_threshold.csv.gz |
212,093,967 | not downloaded | all pairs |
630/no_threshold_connections.csv.gz |
116,904,594 | not downloaded (no header row) | all pairs |
Schema: neurons.csv = root_id, group, nt_type ∈ {ACH,GABA,GLUT,DA,SER,OCT}, nt_type_score, da_avg, ser_avg, gaba_avg, glut_avg, ach_avg, oct_avg (mean per-presynapse probabilities, Eckstein 2024, 6 classes; 9,906 all-zero rows in 783; no confidence floor applied; DA calls drop from 3,189 (630) to 584 (783), reason UNVERIFIED). classification.csv 783 = root_id, flow, super_class, class, sub_class, hemilineage, side, nerve (types live in fw_and_hemibrain_types.csv); 630 also has cell_type, hemibrain_type. super_class vocabulary: optic, central, sensory, visual_projection, ascending, descending, sensory_ascending, visual_centrifugal, motor, endocrine.
Connections: pre_root_id, post_root_id, neuropil (79 values incl. UNASGD), syn_count, nt_type (presynaptic neuron's call). One row per (pre, post, neuropil); the ≥5 filter is on the pair total (per-row counts go to 1). Summing rows recovers exactly 2,613,129 (v630, Lin24) and 2,700,513 (v783, Dork24) pairs. Two synapse detections since July 2025: Buhmann 2021 (connections.csv, connections_buhmann*) and Princeton/Yu 2025 (connections_princeton*, ~48 % more synapses; Codex's live UI count 3,732,460 comes from it). Codex CSVs expose no per-synapse cleft scores; Shiu and Lin used cleft score ≥ 50 on Buhmann synapses.
Shiu model inputs vs Codex v630: 127,319 IDs in common; 659 Codex-only neurons have no ≥5 connection anywhere; 81 Shiu-only IDs are pre-freeze root IDs (UNVERIFIED); every Codex v630 pair appears in the Shiu file with identical weight.
Citations (from Codex's own citation table): always co-cite Dorkenwald et al. 2024 (10.1038/s41586-024-07558-y) and Schlegel et al. 2024 (10.1038/s41586-024-07686-5); plus Zheng 2018 (EM volume), Buhmann 2021 + Heinrich 2018 (synapses before July 2025) or Yu 2025 bioRxiv 10.1101/2025.07.11.664377 (Princeton synapses), Eckstein et al. 2024 (NT), Matsliah et al. 2024 (optic-lobe types), Lin et al. 2024 (connectivity tags), Deutsch 2025 (dsx/fru). For annotation versions ≥3.0.0 also Berg et al. 2026.
Licence: Nature papers and Zenodo deposits CC BY 4.0; the Codex bucket carries no LICENSE object; FlyWire ToS applies to accounts (text UNVERIFIED).
3. Shiu et al. 2024 model inputs (sim golden data)
Repo philshiu/Drosophila_brain_model, MIT. Raw URLs https://raw.githubusercontent.com/philshiu/Drosophila_brain_model/main/<file>.
| File | Bytes | SHA-256 |
|---|---|---|
2023_03_23_connectivity_630_final.parquet |
86,630,944 | 94db8c650533bc36ffa3223f2e62325d5648b8d6bd31c3a4e1c804628c7557b3 |
2023_03_23_completeness_630_final.csv |
3,057,611 | e6b71e17671a9bdb05f55e4bc6774640a1418cb7a05125e0fc994ad40f9bfdfb |
Connectivity_783.parquet / Completeness_783.csv |
100,804,642 / 3,327,347 | not downloaded |
results/example/sugarR{,_100Hz,-<id>}.parquet |
0.5–0.9 MB each | copied to .cache/data/shiu/results/ |
Columns: Presynaptic_ID, Postsynaptic_ID, Presynaptic_Index, Postsynaptic_Index, Connectivity, Excitatory (±1), Excitatory x Connectivity. 127,400 neurons, 14,687,178 edges, min 1 synapse. Cite Shiu et al. 2024 (10.1038/s41586-024-07763-9). Raw model outputs (4.5 GB) on Edmond doi:10.17617/3.CZODIW.
4. hemibrain v1.2 / v1.2.1
- neuPrint
hemibrain:v1.2.1only;Meta: totalPreCount 9,496,606, totalPostCount 64,139,744,postHighAccuracyThreshold0 (weights include low-confidence PSDs;weightHPuses 0.7), no NT properties on Neuron nodes. - Bucket
gs://hemibrain/v1.2/(public):exported-traced-adjacencies-v1.2.tar.gz45,872,577 B, SHA-25607d8946eb0c4e3a5cb23d5769c9817847494f9fcadbc0ca239eed7bbd5555cf7→traced-neurons.csv(21,739 non-cropped Traced;bodyId, type, instance),traced-total-connections.csv(3,550,403 edges, Σ 14,329,229, ≥5: 662,578),traced-roi-connections.csv(62 ROIs incl.NotPrimary);hemibrain-v1.2-body-mean-neurotransmitters.feather45,591,786 B, SHA-256aab49d858415f559f469a9293adfb4d58e423db83a5debb24272ee4d66e059ad(837,710 bodies;gaba, acetylcholine, glutamate, serotonin, octopamine, dopamine, neitherprobabilities summing to 1;predicted_nt; uploaded 2026-05-15, provenance UNVERIFIED, no README); also per-T-bar NT feather (443 MB bz2), all-traced adjacencies, synapse partners (859 MB), skeletons. - v1.2 export vs v1.2.1 DB: 342 edges / 4,077 synapses differ ("minor fixes").
- Side/hemilineage are not in neuPrint; use Schlegel Supp. 5 or sjcabs
hemibrain_121_meta. - Licence CC-BY (Janelia page; version UNVERIFIED). Cite Scheffer et al. 2020, eLife 9:e57443 (10.7554/eLife.57443).
- Converter
flyconn.data.convert.hemibrain(flyconn data pull hemibrain@1.2.1): universe and edges from the export tarball;Supplemental_file5_hemibrain_meta.csvpinned athttps://raw.githubusercontent.com/flyconnectome/flywire_annotations/v3.1.0/supplemental_files/Supplemental_file5_hemibrain_meta.csv(2,634,367 B, SHA-256c9ca8421a5b6a5382348e06a7f62ef452bc78562983deff9ff479e30be2bd2a2; same bytes onmain); sjcabshemibrain_121_meta.feather(2,023,858 B, SHA-256a2775c31450d3a060875bd5690adcb9800d3c9a580a9a919e10829984c79a1d1, MD5 matches the bucket listing). The tarball is registered at levelmetabecause it carries the neuron list. Results in GOLDEN_RESULTS 6e.
5. MANC v1.0 / v1.2.1
- neuPrint
manc:v1.0(23,759 Traced today),manc:v1.2.1(23,665 Traced; 5,305,638 Traced→Traced edges, Σ 30,934,610 = Codex's unthresholded count),manc:v1.2.3(property update).postHighAccuracyThreshold0.4. - Flat export only for v1.0:
gs://flyem-manc-exports/v1.0/(manc-v1.0-neuron-properties.feather17,188,218 B, SHA-2560c4476528906bb0a20e05e1f01e83fc2e5a582761536171ede5463669ca0b891, 102,369 bodies × 56 cols incl.ntGabaProb, ntAcetylcholineProb, ntGlutamateProb, ntUnknownProb, predictedNt, predictedNtProb;traced-neurons.csv23,188;traced-connections.csv; neuPrint CSV/Feather dumps). v1.2 synapse partners:gs://manc-seg-v1p2/manc-v1.2-synapse-partners-minconf-0.0.feather(1.94 GB). Codexgs://flywire-data/codex/data/manc/1.2.1/neurons.csv.gz(23,665 rows, SHA-256ddeea1ff2859941d60a8f24933eb963d1f65d28d2399768c589a74bb90ac4a80). - NT classifier is 3-class (ACh/GABA/Glu) + unknown/unclear; no monoamines. Sides
RHS/LHS; hemilineages Truman0A…27X. - Licence CC-BY. Cite Takemura et al. 2024 eLife 13:RP97769; Marin et al. 2024 eLife 13:RP97766; Cheong et al. 2025 eLife 13:RP96084.
- Converter
flyconn.data.convert.manc(flyconn data pull manc@1.2.1) reads the sjcabs compiled files:compiled_data/manc_121/manc_121_meta.feather(1,446,722 B, SHA-256067c88e687ade67bd1e69b88765b4b21596280618cefde0fa43117d4af4a9439; 23,650 rows = 23,665 Traced minus 15 glia) andmanc_121_simple_edgelist.feather(87,386,906 B, SHA-256c3f9ca6dc9d100e72299b41136d373ffcd60b328624cbfd6ac145f071e0b28ac, MD57ko0TfSZiZ1EGPF2O8OaMA==as listed; 5,305,354 rowspre, post, count, norm, total_input, string ids, Σ count 30,943,884;total_input= Σ count per post over traced partners), plus Codexneurons.csv.gzfor NT confidence andvfbId. sjcabs data are CC BY by the dataset creators; cite the MANC papers. Results in GOLDEN_RESULTS 6e.
6. BANC v888
- Public bucket
gs://lee-lab_brain-and-nerve-cord-fly-connectome/(README, CHANGELOG,compiled_data/,neuron_annotations/v888/,neuron_connectivity/v888/,synapses/,nblast/, meshes, skeletons); Harvard Dataverse doi:10.7910/DVN/7WTH1N v3.0 (2026-07-01, CC BY 4.0, 536 GB). CAVEbrain_and_nerve_cord_public(token). Codexbanc/888/. - Paper: Bates AS, Phelps JS, Kim M, Yang HH, … Lee W-CA, Wilson RI. Distributed control circuits across a brain-and-cord connectome. Nature 656:957–970 (2026). doi:10.1038/s41586-026-10735-w. CC BY 4.0.
banc_888_meta.feather57,503,026 B, SHA-25686ccf5df0c67419f8c5f43e93a7ed38d23a080e9f7fde26737290252f3780098, 188,508 × 81 (patched in place after publication:root_890present; pin MD5). Identifier columns are strings.proofreadTRUE 150,952 +roughly_proofread5,060 = 156,012 (paper: 155,916).super_classBANC vocabulary;neurotransmitter_predicted8 classes (adds histamine, tyramine) withneurotransmitter_score; cross-refsfafb_cell_type(131,240),manc_cell_type,malecns_cell_type(43,490; v0.9),hemibrain_cell_type,fanc_cell_type,*_matchIDs,*_nblast_match.neuron_annotations/v888/codex_annotations.parquet74,350,281 B, SHA-2568a52f5f84f439603881bb360ea990601fd593072261c8cfb9afb0b3842f89eda: long-form, 1,841,078 rows, 31 classification systems incl.fafb_783_cell_type,fafb_783_match_id,manc_121_*,hemibrain_121_*,malecns_09_*,fanc_1116_*.- Edgelists (not downloaded):
compiled_data/banc_888/banc_888_edgelist_simple_v2.feather305 MB (11,510,975 rows; synapse detector v2, size ≥5, no pair threshold),_v3359 MB (13,507,098; new detector, size ≥10). Columnspre, post, count, norm, pre_count, post_count. Per-synapse 8-NT probabilities:synapses/v3.0/banc_nt_prediction_v3_w_sizethresh_10_05042026.parquet5.79 GB. - Codex
banc/888/neurons.csv.gz2,881,728 B, SHA-25640a2201554a8c34d2c4b07c8322543a07c1b3faf5acafbac363fd1a3d0fa617f, 158,262 rows; Codex BANC default threshold 3. - Caveats: both antennal nerves damaged; two synapse versions (paper v2, recommended v3); 22 known problem regions; left optic lobe less typed.
7. FANC
- Public exports
gs://lee-lab_female-adult-nerve-cord/CAVE/{v1237,v1444}/(parquet with CAVE metadata; datastackfanc_production_mar2021):cell_ids_v2.parquet(21,978;user_idis the stable cell ID),somas_dec2022,neck_connective,peripheral_nerves,proofread_first_pass(7,246),proofread_second_pass(1,771),synapses_nov2022_…connectioncounts_countthresh3.parquet(1,878,659 pairs ≥3, Σ 11,801,582; most postsynaptic roots are unproofread fragments), per-neuropil variant (43 neuropils/tracts). SHA-256s in the research note. - Live CAVE is restricted to community members (PI emails the Lee lab; no public applications). No cell-type table in the public export; FANC types reach us only via BANC (
fanc_1116_cell_type, 2,905 neurons). - Licence: none stated (UNVERIFIED). Cite Phelps et al. 2021 Cell; Azevedo et al. 2024 Nature 631:360 (10.1038/s41586-024-07389-x); Lesser et al. 2024 Nature 631:369 (10.1038/s41586-024-07600-z).
8. sjcabs compiled_data (harmonized vocabulary source)
Bucket gs://lee-lab_brain-and-nerve-cord-fly-connectome/compiled_data/{banc_888,fafb_783,manc_121,hemibrain_121,malecns_09,fanc_1116}/; built by flyconnectome/bancpipeline; code MIT; data "CC-BY by the respective dataset creators"; asks to cite original papers + Eckstein 2024. Meta feathers on disk (SHA-256 in the note): fafb_783 144,837 × 24; hemibrain_121 25,397 × 19; manc_121 23,650 × 15; malecns_09 165,114 × 33; banc_888 188,508 × 81. Thirteen columns are common to all five: region, hemilineage, nerve, flow, super_class, cell_class, cell_sub_class, cell_type, neurotransmitter_predicted, cell_function, cell_function_detailed, body_part_sensory, body_part_effector. Edgelists carry pre, post, count, norm[, total_input] with no pair threshold. BANC-space files for other datasets are morphology only.
9. Harmonized schema alignment (seed for M1)
| flyconn field | sjcabs/BANC | Codex FAFB | MaleCNS v1.0 | hemibrain | MANC | Notes |
|---|---|---|---|---|---|---|
neuron_id |
<ds>_id (string) |
root_id int64 |
bodyId |
bodyId |
bodyId |
store as uint64/string; keep source dtype in provenance |
flow |
flow |
flow |
derive from superclass |
– | derive from class |
afferent/intrinsic/efferent |
super_class |
BANC vocab | central, optic, … |
cb_/ol_/vnc_intrinsic, … |
Schlegel Supp. 5 / sjcabs | class |
map to BANC vocab; keep super_class_raw |
cell_class, cell_sub_class |
long snake_case | class, sub_class |
class, subclass |
– | subclass |
keep raw |
cell_type |
cell_type |
cell_type + hemibrain_type (783: separate file) |
type |
type |
type, systematicType |
|
hemilineage |
ItoLee / Truman mixed | ItoLee | itoleeHl, trumanHl |
Supp. 5 | Truman | two columns |
side |
left/right/center | side |
somaSide L/R/M |
Supp. 5 | somaSide LHS/RHS |
normalise to left/right/center |
nt_pred, nt_conf, nt_p_* |
neurotransmitter_predicted/_score (string dtype in F/H!) |
nt_type, nt_type_score, six *_avg |
predicted_nt, predicted_nt_confidence, consensus_nt (no probs) |
feather 6+neither probs | 3-class probs | record class set per source (3/6/7/8) |
| cross-refs | fafb_cell_type, manc_cell_type, hemibrain_cell_type, malecns_cell_type, fanc_cell_type, *_match |
hemibrain_type |
flywireType, hemibrainType, mancType, mancBodyid |
– | synonyms |
plus vfb_id/fbbt_id as neutral keys |
| edges | pre, post, count, norm (no threshold) |
pre_root_id, post_root_id, neuropil, syn_count (pair ≥5, per-neuropil rows) |
body_pre, body_post, weight (no ROI) |
bodyId_pre, bodyId_post, weight (+ROI file) |
neuPrint weight |
store threshold and synapse_source metadata per edge table |
10. Open items
CAVE materialization lists (token); hemibrain NT feather provenance; FlyWire ToS text and the 630→783 DA-call drop; the 81/659 Shiu–Codex ID differences; FANC licence; MANC v1.2.x flat exports; MaleCNS NT "unclear" thresholds and the 11,751 vs 11,710 type delta (Cell Methods).