A working reference for the technical track: landmark datasets with specifications and access routes, the software landscape by workflow stage, benchmarks, and the curation schema for adding entries.
This is a lookup table, not a lesson. Come here when you need to answer one of:
Specifications below are as published and are approximate where releases have been revised. Always confirm current figures and access terms against the primary source before citing them — dataset sizes in particular change as proofreading continues (Unit 04 §2).
| Dataset | Scale | Modality / resolution | What it is good for | Access |
|---|---|---|---|---|
| C. elegans (White et al. 1986; Cook et al. 2019; Witvliet et al. 2021) | 302 neurons; whole animal; a developmental series across maturation | ssTEM | The complete-nervous-system reference case. Witvliet’s series is the best available data on how connectivity changes with development | WormWiring, WormAtlas |
| Larval Drosophila brain (Winding et al. 2023) | ~3,000 neurons; ~550,000 synapses; whole brain | ssTEM | A whole brain small enough for exhaustive graph analysis; bilateral matching studies | CATMAID instances; published supplements |
| Drosophila hemibrain (Scheffer et al. 2020) | ~25,000 neurons; ~20 million synapses; central brain | FIB-SEM, near-isotropic 8 nm | Cell-type census; the cleanest large connectome for analysis teaching, because isotropy makes tracing quality high | neuPrint |
| FAFB / FlyWire (Zheng et al. 2018; Dorkenwald et al. 2024) | Whole adult brain; ~139,000 neurons; ~54.5 million synapses | ssTEM, 4 × 4 × 40 nm | The first whole-brain connectome of a behaviorally complex animal; community-proofread | FlyWire (registration required); CAVE |
| Drosophila male adult nerve cord (MANC) | ~23,000 neurons | FIB-SEM | Motor circuits; connecting brain to periphery | neuPrint |
| Dataset | Scale | Modality / resolution | What it is good for | Access |
|---|---|---|---|---|
| Mouse retina (e2198 and relatives) (Briggman, Helmstaedter, Denk) | ~10⁵–10⁶ µm³ | SBEM | Structure-function in a well-characterized circuit; direction selectivity | Published; some via community portals |
| Kasthuri saturated reconstruction (Kasthuri et al. 2015) | ~1,500 µm³ mouse neocortex, densely reconstructed | ssSEM (ATUM) | The reference for what dense, saturated reconstruction means and costs | Open Connectome / BossDB |
| Hippocampal CA1 resource (Harris et al. 2015) | Dense neuropil volume | ssTEM | Spine and synapse ultrastructure; a standard teaching set for Units 05–06 | Published resource |
| MICrONS (Allen Institute, Baylor, Princeton; 2025 release) | ~1 mm³ mouse visual cortex; ~200,000 cells; ~500 million synapses | ssTEM, 4 × 4 × 40 nm, co-registered with in-vivo two-photon calcium imaging | The reference functional-connectomics dataset. The co-registration is what makes it unique | CAVE / caveclient; MicronsBinder notebooks |
| H01 human cortex (Shapson-Coe et al. 2024) | ~1 mm³ human temporal cortex; ~57,000 cells; ~150 million synapses; ~1.4 PB | ssTEM, ~4 × 4 × 30 nm | Human tissue at synapse resolution; species comparison | Google/Lichtman lab public release; Neuroglancer |
| Larval zebrafish whole brain (Hildebrand et al. 2017) | Whole brain, larval | ssEM | Whole-vertebrate-brain scale in a tractable organism | Published resource |
| MouseConnects / HI-MC (NIH BRAIN CONNECTS) | Scaling toward whole mouse brain | Volume EM | The current flagship scaling effort; see the case study | Program resources |
| If your question is about… | Start with |
|---|---|
| Whole-brain graph structure in a behaving animal | FlyWire, or larval Drosophila for exhaustive analysis |
| Cell types and their connectivity, with clean tracing | Hemibrain |
| Structure–function relationships | MICrONS |
| Human-specific features | H01 |
| Development of connectivity | C. elegans developmental series |
| What dense reconstruction actually requires | Kasthuri 2015 |
| Teaching ultrastructure reading | Harris CA1 resource; any of the above in Neuroglancer |
Before proposing new acquisition, check this table. A large fraction of good connectomics questions can be answered by re-analysis of existing public data. The cost difference is not marginal: acquiring a new mm-scale volume is a multi-year, multi-million-dollar program, while re-analyzing one is a compute bill and your time. Work the comparison out for your own question with Unit 03’s cost arithmetic before assuming you need new data. See Unit 02’s common errors.
| Stage | Tool | What it does |
|---|---|---|
| Viewing | Neuroglancer | The standard browser-based viewer for petascale volumes, meshes, and annotations |
| webKnossos | Viewing, annotation, and proofreading with collaborative features | |
| Data access | CloudVolume | Python access to precomputed/chunked volumes |
caveclient |
Client for CAVE — segmentation, synapse tables, materialization versions | |
neuprint-python |
Client for neuPrint (hemibrain, MANC) | |
intern |
Client for BossDB-hosted volumes | |
| Storage formats | Precomputed, N5, Zarr / OME-Zarr | Chunked, multiresolution array formats (Unit 04 §3) |
| Annotation / proofreading | CATMAID | Collaborative skeleton tracing; the system behind FAFB-era tracing |
| VAST | Manual volume annotation | |
| CAVE / ChunkedGraph | Versioned, concurrent proofreading over immutable supervoxels | |
| Segmentation | Flood-filling networks | Iterative single-object growth |
Affinity + agglomeration stacks (e.g. gunpowder-based pipelines, PyTorch Connectomics) |
The dominant production family | |
| Ilastik | Interactive pixel classification for smaller volumes and prototyping | |
| Morphology | navis, natverse |
Neuron morphology analysis in Python and R |
meshparty |
Meshes and skeletons from segmentations | |
| Graph analysis | networkx |
General graph analysis, triad census |
graspologic |
Statistical graph analysis: embeddings, graph matching, two-sample testing | |
| DotMotif | Declarative subgraph/motif queries over connectomes | |
| Hosting | BossDB | Community archive for volumetric neuroscience data |
| Benchmark | What it evaluates | Notes |
|---|---|---|
| SNEMI3D | 3D neurite segmentation | Long-standing reference challenge; small volume |
| CREMI | Neuron segmentation and synaptic partner identification in Drosophila EM | Includes synaptic partner assignment, which most benchmarks omit |
| ISBI 2012 | 2D membrane segmentation | Historical; useful for teaching, not representative of current difficulty |
Caution when reading benchmark results. Scores on small, well-prepared benchmark volumes systematically overstate performance on production data, which contains artifacts (Unit 03), rare morphologies, and volume boundaries that benchmarks exclude. When evaluating a method for your project, the question is not its leaderboard position but its error rate on your tissue — which means running it on a representative sub-volume of your own data (Unit 03 §3, the pilot reconstruction rule).
Report metrics as described in Unit 08 §3: at minimum VI decomposed into split and merge components, plus one tracing-oriented metric such as ERL, plus the effect on your endpoint.
Every atlas entry carries:
citation — standardized citation string, with DOIworkflow_stage — acquisition |
reconstruction |
proofreading |
analysis |
cross-cutting |
species, brain_regionmodality, effective_resolutiondataset_or_code_access — URL plus access constraints (open / registration / request)maturity — concept |
validated prototype |
production-validated |
known_limits — one concise sentence on the technical boundarymapped_units — technical-track unit slugs this supportsThe known_limits field is the one most often left blank and the one most worth
filling. An atlas of tools without limits is advertising.
This page is a lookup table and this panel is deck context, not reference material — nothing here should be cited. Use it instead to rehearse the habit §1 asks for: confirm every specification against the primary source, because dataset sizes and cell counts change as proofreading continues.
Module14 L3 S03-02: The opening of the source deck’s reference stream. Treat any figure it carries as provisional — volume sizes, cell counts, and synapse counts in this field are release-dependent, so check them against the primary source before citing (Unit 04 §2).
Module14 L3 S10: Mid-stream references. Use it to test the curation standard in §4: for each resource named, ask whether you could write its known_limits sentence. That field is the one most often left blank, and an atlas of tools without limits is advertising.
Module14 L3 S19: Closing references. Check them against the choosing table in §1 before proposing new acquisition — a large share of good connectomics questions can be answered by re-analysis of existing public data, at a small fraction of the cost.
Techtalk S44: Developmental motif comparison, cross-referenced here from Unit 09. Use it as the pointer into the C. elegans developmental series in §1 — the dataset to reach for when the question is how connectivity changes with maturation rather than what it is in one adult.
Attribution: module14 lesson3 and neuroAI source decks (historical/context visuals).
Add one resource to the atlas. Produce:
Rubric. Proficient: all fields present and accurate. Strong: the limitation sentence is specific enough that a reader could predict a failure case from it, and the unit mapping identifies where in that unit’s workflow the resource fits.
§4 of this page requires every contributed entry to carry “a standardized citation string, with DOI”. This section is that rule applied to the page itself. Every dataset named in §1 is here, with the published reference rather than its preprint — several of these circulated as bioRxiv preprints for years and are still cited that way.
MouseConnects / HI-MC is an in-progress BRAIN CONNECTS program rather than a released dataset; see the case study for its current state. Tool entries in §2 link to their own documentation, which is the citable source for a version.
This unit is anchored to canonical papers and datasets used in connectomics practice. Use these as required preparation before activities.
Capability target: Curate a trustworthy connectomics reference atlas with transparent metadata and evidence quality signals.
Atlas Entry Sprint - Produce publishable reference entries with complete metadata and quality notes. The unit's own lab above is the graded version of this exercise; do that one.
Curate methods, datasets, and tools with metadata completeness and explicit limitations.