SpaMTPData

SpaMTPData >= 0.99.5 attaches declared species and resource provenance to loaded native experiments and their nested altExp() objects in metadata(x)$SpaMTPData. Species checks therefore survive extracting a paired transcriptome, while explicit child species declarations are retained. SpaMTP >= 0.99.5 can use these fields when annotating genes with annotateGeneIdentifiers(), whose HGNC reference is provided by SpaMTPdb >= 0.99.4. Human HGNC mapping is rejected for experiments declared as mouse; it does not perform orthology conversion. The published experiment files, their hashes and resource versions are preserved.

SpaMTPData provides ExperimentHub access to the datasets used by SpaMTP tutorials and tests. SpatialExperiment datasets, Cardinal experiments and auxiliary files remain outside Git and are downloaded only when requested. ExperimentHub is preferred; before Hub ingestion is complete, the same API falls back to immutable Zenodo URLs with a 30-minute timeout, retries, and size/MD5 validation.

# Native, deterministic synthetic example: no download or Seurat dependency.
spe <- SpaMTPData::spaMTPExampleData()
SummarizedExperiment::rowData(spe)$mz
SingleCellExperiment::altExpNames(spe)

# Inspect a published resource before downloading its original representation.
SpaMTPData::spaMTPData("mouse_brain_dhb_striatum", metadata = TRUE)

The coordinated API uses spaMTPData(), spaMTPDataResource(), spaMTPDataResources() and spaMTPDataVersion(). This replaces the historical PascalCase exports. Use SpaMTPData >= 0.99.4 with SpaMTP >= 0.99.4 and SpaMTPdb >= 0.99.3 for the coordinated release. Package versions are distinct from resource releases 1.1.0 and 1.0.0.

The default resource release, 1.1.0, provides seven native SpatialExperiment datasets from Zenodo record 22733262. The historical 1.0.0 release remains available explicitly; its seven Seurat entries retain requires_conversion = TRUE, their original URLs and hashes. Historical objects are not silently converted or relabelled.

The one-time native preparation recipe extracts the required data from all seven archives into standalone SpatialExperiment RDS. The published 1.1.0 files retain counts/data, annotations, source-image centroids and paired transcriptomes, and can be used directly:

spe <- SpaMTPData::spaMTPData("mouse_brain_dhb_striatum", version = "1.1.0")
SingleCellExperiment::altExpNames(spe) # SPT

Preparation needs optional Seurat support once; reading these native RDS does not. The prepared files have their own provenance and checksum manifest; they are not uploaded by the recipe or silently substituted for old published URLs.

For offline development, set options(SpaMTPData.resource_dir = "...") to a directory containing files named as listed in the resource registry. A version-specific subdirectory is preferred. Local files are checked against published sizes/checksums by default; verify = FALSE is for deliberate development fixtures only, and never disables class validation. Verified source-download caches also work with offline = TRUE, without a Hub connection. Cache entries are versioned to avoid filename collisions.

Each release contains 18 resources: the current native snapshot and the historical snapshot together have 36 registry entries, covering mouse brain, mouse bladder, import examples, simulated single-cell multi-omics, annotation refinement, a large-data ROI, and the human brain demonstration.

SpaMTPData owns experiment resources, SpaMTPdb owns annotation tables and SpaMTP owns analysis/conversion. Neither data package imports SpaMTP or Seurat. Physical files share the existing SpaMTPdb Zenodo version family; this does not require a separate SpaMTPData deposition family or a dependency between the two R packages. The shared publication recipes live in SpaMTPdb >= 0.99.3. The native synthetic example is separate from the published-resource registry; it is a teaching fixture, not an additional biological dataset.

After a shared native release is publicly available, run the registration recipe from a SpaMTPData >= 0.99.3 checkout:

Rscript inst/scripts/register-native-release.R \
  /path/to/shared-staging PUBLISHED_RECORD_ID /path/to/new-registration-output

This rejects draft IDs and other Zenodo version families, verifies file sizes and MD5 against the prepared native manifest, and reads all seven public files in a fresh temporary cache as SpatialExperiment without Seurat. It writes candidate resource_manifest.csv and metadata.csv files to the specified output directory; it does not modify this package's live registry or ingest Hub records. Review and validate the candidate tables before installing them in inst/manifest and inst/extdata, respectively.

The new experiment release is a complete snapshot: seven native replacements plus the eleven unchanged resources at their existing immutable locations. Historical 1.0.0 entries remain available. Publishing a new resource does not silently replace the source class or checksum of an older release.

If you use these resources, cite the SpaMTP paper: Causer, Lu, Kriel et al., Nature Methods (2026).