Reproduce the examples
The eleven public example apps are live in the examples gallery. They are built from https://github.com/MicTott/spatialscape-examples: one folder per dataset with download scripts, any required preparation scripts, a dataset.yaml, and a README documenting the data source, license and resource requirements. Check the dataset's README before running its build; larger examples require a cluster.
| Folder | Technology | Scale | Source | Live app |
|---|---|---|---|---|
visium-human-dlpfc | Visium, H&E, manual layer labels | 12 sections, 47,681 spots | spatialLIBD (Maynard et al. 2021) | open |
visium-cytassist-mouse-brain | Visium CytAssist, H&E | 1 section, 2,310 spots | 10x Genomics | open |
visium-hd-mouse-brain | Visium HD, 8 µm bins, H&E | 1 section, 394k bins | 10x Genomics | open |
xenium-mouse-brain | Xenium Prime 5K: image, shapes, nuclei, transcripts | 1 section, 63k cells | 10x Genomics | open |
atera | Atera whole-transcriptome in situ: morphology, shapes, transcripts | 2 sections, 888k cells, 1.66 billion transcripts | 10x Genomics previews | open |
merscope-mouse-brain | MERSCOPE: boundaries, transcripts, mosaics | 3 slices, 233k cells | Vizgen | open |
cosmx-nsclc | CosMx SMI: cell types and niches, boundaries, transcripts, composites | 8 sections, 800k cells | NanoString (He et al. 2022) | open |
phenocycler-lymph-node | PhenoCycler spatial proteomics: 56-channel image, cell outlines, gated cell types | 3 cores, 46,671 cells, 56 proteins | BioImage Archive S-BIAD2100 | open |
spatial-atac-rna-mouse-brain | spatial ATAC + RNA: RNA on 20 µm pixels, ATAC tracks per cluster | 3 sections, 14k pixels | GEO GSE205055 | open |
pbmc-multiome | 10x Multiome: RNA and ATAC UMAPs of the same nuclei, ATAC tracks per cell type | 11,577 nuclei | 10x Genomics | open |
abc-atlas | whole mouse brain: 59 MERFISH sections and all 10x nuclei on Allen's UMAP | 4.0 M cells + 4.0 M nuclei | Allen Brain Cell Atlas | open |
Start with an R object
The visium-human-dlpfc example reads the spatialLIBD SpatialExperiment directly. Its build.sh downloads the object, generates a configuration, builds the committed configuration and validates the result. It follows the workflow in Getting started:
git clone https://github.com/MicTott/spatialscape-examples && cd spatialscape-examples
bash setup_env.sh && source env/bin/activate # the spatialscape release the apps were built with
cd visium-human-dlpfc && bash build.sh # download.R, init, build, validate
spatialscape serve bundles --openCompare dataset.init.yaml with the committed dataset.yaml after the run to see which annotations and display settings the example uses. The download step requires R with spatialLIBD installed.
Build from published inputs
Each dataset's README lists its download and preparation steps. For example, from the repository root with the environment active:
cd xenium-mouse-brain
bash download.sh
python prepare.py
spatialscape build dataset.yaml
spatialscape serve bundles --openSome datasets read the downloaded object directly and do not have a preparation script. The YAML also references images, boundaries or transcript tables under raw/. If an optional file is absent, the build reports it and omits that layer; downloading the file and rebuilding adds it.
Processed-data downloads
The repository includes fetch_processed.sh for downloading prepared objects from Zenodo. Its default record ID is currently a placeholder, so this route is not yet available. Use the published-input workflow above until the repository provides a record ID and DOI. The Allen atlas is excluded from the planned processed-data archive because of its size.
On a cluster
Each folder's slurm/ has the same steps as SLURM jobs with the memory and time each one needed; the README of the repository says how to chain them. The Allen atlas is the one dataset that is cluster-scale (215 GB of processed objects, about a day of jobs) and is not on Zenodo; its raw data are Allen's own public bucket.
Runtime, memory and storage
The dataset READMEs report runtime, peak memory and disk use for the machines or cluster jobs used to build the examples. Use these reports to estimate requirements for your run. Benchmarks describes the build and viewer measurements and how to collect them on your own machine.