DigitalBrain Data Explorer

Developed by LattixOrigin and Lu Meng lab @PKU

格原万维 LattixOrigin
About DigitalBrain

DigitalBrain brings human brain single-cell datasets into a unified atlas. Explore gene expression, cell types and brain regions across studies.

DigitalBrain workflow from a whole-brain atlas with anatomical and lifespan coverage, through DigitalBrain-M1 gene and cell embeddings, to hierarchical functional connectivity and ageing analysis of dentate gyrus granule cells.

DigitalBrain_Atlas integrates large-scale human single-cell transcriptomic datasets across broad anatomical regions and the human lifespan into a harmonized whole-brain atlas with standardized anatomical and cellular annotations, creating a unified, machine-readable resource for AI modeling. This atlas provides the foundation for DigitalBrain-M1, a human brain foundation model that learns gene- and cell-level representations directly from the integrated data.

These representations recover higher-order biological organization, including hierarchical relationships among brain regions and patterns aligned with functional connectivity, indicating that the model captures structure beyond raw gene expression.

As a biological stress test, ageing analysis identifies the dentate gyrus as a particularly vulnerable region and reveals age-associated remodeling at multiple levels, including gene-regulatory networks and gene/cell-state manifolds. Together, the framework illustrates how a comprehensive, harmonized and AI-ready human brain atlas can move beyond data aggregation to support representation learning, systems-level organization and discovery of biologically meaningful changes across the human brain.

View full-size PDF

Cite DigitalBrain

If you use DigitalBrain, the Data Explorer, or data processed through this resource in your research, please cite:

An J, Hu X, Jiang Y, et al. Multiscale transcriptomic organization of the human brain with DigitalBrain. bioRxiv (2026). 2026.04.14.718492. Preprint

View preprintFull citation details

Please also cite the relevant primary publications when using individual datasets. Data sources & acknowledgements

Gene expression is a cross-study merge with the dataset axis collapsed, so it is always shown at global scope. Collection, dataset and donor filters do not apply to this layer.

Current Scope

All Collections

Use the filters above to move from collection-level summaries down to donor-specific views.

The 3D atlas is shown first. Switch to Overview here for cell-type and brain-region charts.

Datasets 0
Donors 0
Cells 0

Collection Overview

Datasets 0
Donors 0
Total Cells 0
Brain Regions Brodmann 0 Gyral 0
Cell Types 0

Dataset Overview

Donors 0
Total Cells 0
Brain Regions Brodmann 0 Gyral 0
Cell Types 0

Donor Details

-
Age: - Gender: -
0 cells

Cell Type Distribution

Composition, diversity, and comparison views for the current scope.

Brain Region Distribution

Composition, coverage, and comparison views for regional annotations.

Please select a collection to view data
Interactive 3D Atlas
Whole-brain cells + Allen 3D anatomy 105 profiles placed
105 31 cell types

Age / developmental stage

Custom postnatal range, in years

Ranges include records whose reported age falls entirely within the bounds. Prenatal stages use a separate time basis.

Donor status

Source-reported conditions in the harmonized classification. A parent includes its subtypes.

No matching conditions. Try another name or abbreviation.

Sample tissue status Optional

Use for explicitly annotated sampling contexts, such as tumour tissue. An unannotated sample is not assumed normal.

Loading cohort filters...

No cells match these filters. Broaden the age range, remove a condition, or include unknown annotations.

Age and status filters are not yet supported in this view. Support is planned for a future update.

Choose cell types in .

Cell type
Cell type

Tick any number of classes: the atlas is recoloured from those alone, not just the detail panel. “only” isolates one class.

0%
0%100%
96%
0%100%

55 regions shared by the FC and SC matrices

Label DMN nodesCore and associated nodes in this atlas vocabulary
3D anatomyOfficial Allen voxel geometry
Region envelopesFilled parcel boundary overlays
Auto-rotatePause automatically on selection
i

Anatomy is derived from the Allen Human Reference Atlas – 3D, 2020. When linked to a selection, only sampled regions are shown (sized by cell count); composition is scope-level (marginal), not region-resolved.

Functional focusMemory—
No regions match the current filters.
SA
Excitatory neuron12–52%
LowerRelative abundanceHigher
All cell typesWithin-region composition
Gene expressionMean expression
SparserLit point densityDenser
Functional connectivityTop 4%
P96Within-matrix percentileP100
≥ 0.3893.800
— links
DMN nodeFunction-focused link
Anatomy parcelParcel boundaryCell profile
Exact anatomyCoarse ontology proxyCurated gyral proxy
Drag to rotate anatomy · Scroll to zoom · Click a cell-profile marker

Select a region

Select a mapped parcel to inspect its anatomical support and cell-class profile.

—

—

—

Excitatory neuron—
—

Region role and functions

Curated context

—

—

Selected genes here

Mean expression

Cell-class composition

Relative proportion

Member regions

—

Strongest connections

Weight

Selected connection

FC
—
fMRI FC weight
—
Matrix percentile
—

—

Descriptive association only; no direction or causality is implied.

Atlas context

Region universe
Table 2 hierarchy
Spatial layout
Allen 3D mapping
Mapping basis
—
Data status
Illustrative

Across brain regions

—

Housekeeping genes are drawn as reference bars, never as part of the selection. Classic controls (ACTB, GAPDH) are the familiar ones but vary 2.5-2.8x between brain cell types; the stable set varies 1.4-1.7x and is the one to trust when the levels matter. In the type panel each control is read in that type, so it shows how low the baseline sits there before any gene is compared against it.

Within cell types

—

Data and interpretation

A scientifically cautious prototype

Region identities come from the 106-region DigitalBrain hierarchy. When linked to a Data Explorer selection, sampled regions are matched on the Brodmann axis and sized by real per-region cell counts.

Cell-type composition uses the Data Explorer's original taxonomy. Because the source data stores only marginal distributions (region totals and cell-type totals, not a joint region × cell-type matrix), the per-region composition shown is scope-level, not region-resolved.

Functional and structural connectivity are read from the project's 55-region FC and SC matrices; links are undirected and descriptively thresholded by matrix percentile.