ηJump

Applied to the cells embedded in this page, mostly entirely in your browser. Leave a section untouched to ignore it. The one exception is current community identity, below, which reads the community's own reports the first time you open this tab.

A cell matches if it falls inside any box. Each box is drawn in its own colour in the Neuroglancer link, on its own toggleable layer.

The dataset. Shapson-Coe, Januszewski, Berger et al., “A petavoxel fragment of human cerebral cortex reconstructed at nanoscale resolution,” Science 384, eadk4858 (2024). Released CC BY. Every cell type, cortical layer and segment ID in this tool comes from that release.

Volumes and browser: H01 release page. This tool reads the c3 segmentation and its segment_properties, the cell_bodies centroids, and the layers volume.

Cortical layers. Each cell’s layer is sampled from H01’s own layers volume. H01 publishes no mapping from its numeric labels to layer names, so the mapping (1 = L1 … 7 = WM) was recovered here by cross-tabulating against the classification’s #L1…#WM tags: 94.8 % agreement, and every disagreement is exactly one layer apart, never two.

For whole-cortex comparison: Wagstyl et al., “BigBrain 3D atlas of cortical layers: cortical and laminar thickness gradients diverge in sensory and motor cortices,” PLOS Biology 18, e3000678 (2020). It reports thickness as regional gradients (1.67–4.5 mm total) and shares no coordinate frame with H01, so it can sanity-check this sample’s ~2.3 mm but cannot place a boundary inside it.

Cell identification. The guided decision tree is shared with µJump. Perisomatic ultrastructure as a classifier: Elabbady et al., “Perisomatic ultrastructure efficiently classifies cells in mouse cortex,” Nature 640, 478–486 (2025).

Brain ultrastructure overview: “Brain Ultrastructure: Putting the Pieces Together,” PMC7930431.

Neurovascular volume EM: “Public Volume Electron Microscopy Data: An Essential Resource to Study the Brain Microvasculature,” Frontiers in Cell and Developmental Biology 10, 849469 (2022).

Synapse type I/II framework: Gray, E.G., “Axo-somatic and axo-dendritic synapses of the cerebral cortex: an electron microscope study,” J. Anat. 93, 420–433 (1959) — foundational, no open-access link.

The reference images in the guided tree are the MICrONS mouse visual cortex library. Organelle ultrastructure is broadly comparable to human; cell and vessel dimensions are not.

Read these before trusting a number. spiny-stellate (190 cells) sits 62.6 % in L5 rather than L4 by the layer volume, and has the largest median soma of any type — treat it as uncertain. 1,468 c3 segments contain more than one soma, so a segment ID is not always one cell. A further 1,917 detected somata carry no cell-type call at all and are not loaded here. Cell contacts are mesh surface proximity, not synapses — H01 publishes its own synapse tables, which this tool does not yet use.

Fifteen charts computed from the cells embedded in this page — no network. Three more come from the community backend.

Colour is never the only channel. Where a chart is grouped by cell type, the type is on the labelled axis and in the hover text and the hue only repeats it — using the same palette as µJump, δJump and πJump, so a cell type is the same colour in every tool. Where colour carries something on its own it is broad class (three slots) or magnitude through lightness, both of which survive colour-vision checks.

Layer is H01's own layers volume where it covers the soma, otherwise the published tag.

Each panel is one type's own depth distribution, as a percentage of that type — so a rare type is still readable next to a common one. Depth is continuous, measured from the L1/L2 boundary; dotted lines are the fitted layer boundaries. This is the chart µJump cannot draw, because it only has discrete layer bins.

Of every synapse arriving on cells whose soma sits in that layer, the percentage classed excitatory.

From each segment's own voxel count, so it is known for every cell rather than only for cells someone has computed a mesh volume for.

H01's published per-segment spininess — the classic spiny/aspiny discriminator the guided tree asks you to judge by eye.

One point per cell, both axes logarithmic. Coloured by broad class only.

Each row is a cell type; the row shows what percentage of those cells have each type as their nearest neighbour.

H01's layer volume against the classification's own #L1…#WM tags. If the disagreements really are boundary cells, they should pile up at the dotted lines and be near zero between them.

From the shared “Master cell list”, not from H01's frozen call: this one moves when somebody re-identifies a cell. The chart above it is the published classification.

The same cells as the heat map above, but under their CURRENT identity rather than H01's published one — so the two disagree exactly where somebody has corrected a call. Layer is fixed geometry (H01's layers volume where it covers the soma, otherwise the published tag); which type counts in each layer is live.

Split out so the identified types are not crowded by the unclassified population. Same layer estimate.

Every chart above as a spreadsheet: one sheet per cell type, one for type × layer, one for the E/I balance, and one for whatever the live backend last returned.