Omics3D: three-dimensional reconstruction
and analysis of spatial omics
Explore spatial transcriptomics, proteomics and metabolomics in three dimensions. Browse public datasets, reconstruct molecular distributions and analyze your own data with published methods.
Know moreOmics3D Dataset Browser
From files to interpretable results
Four steps for a reproducible spatial workflow.
01Prepare and import your data
Measurements, observation IDs and coordinates
Prepare and import your data
Measurements, observation IDs and coordinatesProvide an observation-by-feature matrix and coordinates with matching observation IDs. Include feature names, source section labels and measurement units. H5AD, NPZ, CSV and TSV imports are supported.
- For H5AD, store coordinates in
obsm["spatial"]. - For NPZ, include
xyz,X,feature_namesandsections. - For tabular files, use one matrix and one coordinate table linked by ID.
02Check coordinates and section geometry
Units, registration and physical spacing
Check coordinates and section geometry
Units, registration and physical spacingReview the import checks before opening a 3D view. Confirm XY units, section order and physical Z spacing. A shared coordinate frame requires author registration or a documented alignment method.
- Use documented acquisition spacing. Display spacing is not instrument resolution.
- Keep independent specimens separate and retain original coordinate provenance.
- Unregistered sections can be inspected individually in the source browser.
03Choose and configure a method
Inputs, parameters and validation
Choose and configure a method
Inputs, parameters and validationIn Analysis, select the dataset, task and method. Review the input summary and required annotations, then adjust the parameters for your sampling geometry and molecular modality.
- Use measured observations for tasks that require measured input.
- Distinguish training, parameter selection and held-out evaluation.
- Review an assistant-prepared configuration before submitting the job yourself.
04Explore and interpret the result
Measured, predicted and author-resampled data
Explore and interpret the result
Measured, predicted and author-resampled dataOpen a completed job to inspect its methods, parameters and outputs. In the viewer, compare molecular abundance with the source observations and examine linked virtual sections.
- Measured marks a source observation or an occupied source voxel.
- Predicted marks a model estimate at an unmeasured position.
- Author-resampled marks an interpolated or resampled author export.
- Review support masks, error units and validation scope before interpreting a pattern.
3D Viewer
Choose a dataset to begin.
Section
Adjust position or thickness
Search datasets
Public datasets
Transcriptomics literature coverage 2016–2026
Dataset details
Spatial preview
Data access
Quality and coordinates
Dataset details
Source processing
Quality checks
—Spatial distribution
Source coordinatesCoordinate and display notes
—
Source coordinates are preserved.Display sampling does not change the source data.—Upload
My uploads 0
Import spatial omics data
File formats and examples
Download CSV example · Extract and upload the matrix and coordinate files. Example units: mm.
Observations × features, with pixel or cell IDs.
id, GeneA, GeneB, …Match matrix IDs and specify coordinate units.
id, x, y, z, sectionSection order, thickness and feature names. Tissue images are optional.
section, z, thicknessH5AD: obsm["spatial"] stores XYZ. NPZ: xyz, X, feature_names and sections. Keep measured and predicted observations distinct.
Optional annotations: H5AD obs["cell_type"], var["gene_symbol"], var["kegg_id"], uns["species"]. CSV coordinates may include cell_type or label. Add a feature table with feature_id, gene_symbol and/or kegg_id for exact ID matching. NPZ accepts cell_types, gene_symbols, kegg_ids and measured flags.
Differential analysis: provide sample_id and condition in H5AD obs or observation_metadata.csv (id, sample_id, condition). Sample IDs must identify independent biological specimens. Cell composition: upload reference_signatures.csv with feature_id and one count-scale signature column per cell type. Feature IDs must match the spatial matrix. Sample metadata template · Reference template · Format guide
Analysis
3D reconstruction
Method details & references
Reconstruct between observations
My Jobs
No jobs yet
Upload a dataset to start an analysis.
Your results appear here
Select a job to view methods, parameters and outputs.