NASA Mouse Spaceflight Models

This repository collects project-specific workflows for mouse spaceflight gene expression modeling. It is no longer only a GLARE workspace: the current repo also tracks OSDR/TMS preprocessing, expiMap/scArches setup, Reactome mouse pathway architecture files, ARCHS4 mouse resources, and downstream analysis.

Repository Layout

• src/nasa_mouse_glare/: project code for OSDR/TMS preprocessing, GLARE adaptation, Reactome GMT generation, and analysis utilities.
• src/glare/: vendored GLARE source with local runtime fixes.
• src/expiMap_scarches/: expiMap/scArches source and handoff notes.
• src/onto-vae/: vendored OntoVAE source used for the parallel pathway-VAE evaluation.
• data/pathways/reactome_current_mouse_ensembl.gmt: generated Reactome mouse GMT file for the expiMap architecture mask.
• assets/archs4/mouse_gene_v2.5.h5: local ARCHS4 mouse H5 resource; ignored by git because it is large.
• data/osdr_api/: NASA OSDR Biological Data API metadata and small manifests; downloaded count CSVs under data/osdr_api/counts/ are ignored.
• literature.md: links for GLARE, VEGA, expiMap, OntoVAE, and MOBER.
• docs/osdr_api.md: NASA OSDR Biological Data API notes and examples.

Setup

Run workflow commands from the repository root:

cd path/to/nasa-mouse
conda activate nasa-mouse
export PYTHONPATH=src

To create or refresh the local environment:

conda create -y -n nasa-mouse python=3.11
conda run -n nasa-mouse python -m pip install -r requirements-nasa-mouse-glare.txt

Current Inputs

• OSDR mouse bulk RNA-seq FLT/GC metadata and count tables are discovered from the NASA OSDR Biological Data API, not from the older local integrated OSDR HDF5.
• ARCHS4 mouse gene expression H5: assets/archs4/mouse_gene_v2.5.h5
• Reactome mouse expiMap architecture GMT: data/pathways/reactome_current_mouse_ensembl.gmt

The Reactome GMT is generated from official current Reactome files:

• ReactomePathways.txt
• Ensembl2Reactome_All_Levels.txt

Regenerate it with:

PYTHONPATH=src python src/nasa_mouse_glare/build_reactome_mouse_gmt.py

The output GMT uses one row per mouse Reactome pathway:

R-MMU-73857_RNA_POLYMERASE_II_TRANSCRIPTION    https://reactome.org/PathwayBrowser/#/R-MMU-73857    ENSMUSG...

Discover OSDR mouse bulk RNA-seq Space Flight/Ground Control samples:

PYTHONPATH=src python -m nasa_mouse_glare.fetch_osdr_mouse_transcriptomics

Audit and prepare API-native multi-tissue GLARE inputs:

PYTHONPATH=src python -m nasa_mouse_glare.multi_tissue_api_glare audit

PYTHONPATH=src python -m nasa_mouse_glare.multi_tissue_api_glare prepare \
  --tissue all \
  --download-counts \
  --prepare-per-study

Outputs are written under outputs/glare_multi_tissue_api/. Retina is audited but skipped for GLARE unless a matching TMS FACS retina pretraining source is added. Skeletal-muscle subtype runs use official OSDR material-type labels and the available TMS FACS limb muscle pretraining tissue.

Run one prepared aggregate scope:

PYTHONPATH=src python -m nasa_mouse_glare.multi_tissue_api_glare run-glare-scope \
  --scope-dir outputs/glare_multi_tissue_api/liver/aggregate

Run MOBER-corrected aggregate GLARE for a multi-study scope:

PYTHONPATH=src:src/MOBER python -m nasa_mouse_glare.multi_tissue_api_glare run-mober-scope \
  --scope-dir outputs/glare_multi_tissue_api/liver/aggregate

Run all per-study GLARE scopes for one tissue and compare against per-study DESeq2:

PYTHONPATH=src python -m nasa_mouse_glare.multi_tissue_api_glare run-per-study-glare \
  --tissue-dir outputs/glare_multi_tissue_api/liver

PYTHONPATH=src python -m nasa_mouse_glare.multi_tissue_api_glare run-dgea-comparison \
  --tissue-dir outputs/glare_multi_tissue_api/liver

Run the paper-style validation stack on the generated multi-tissue GLARE outputs:

PYTHONPATH=src /opt/anaconda3/envs/nasa/bin/python -m nasa_mouse_glare.multi_tissue_validation \
  --include-per-study \
  --include-mober \
  --shap-aggregate

This writes XGBoost verification, representation QC, clustering QC, DEG-enrichment comparisons, intersection-vs-GLARE-only module-score validation, Panglao marker enrichment, and Metascape-ready gene lists to outputs/glare_multi_tissue_api/validation_stack/.

Prepare tissue-specific expiMap inputs from API count tables:

PYTHONPATH=src python -m nasa_mouse_glare.prepare_expimap_osdr_tissue --tissue liver
PYTHONPATH=src python -m nasa_mouse_glare.prepare_expimap_osdr_tissue --tissue kidney

Workflows

The GLARE-compatible preprocessing and fine-tuning workflow is documented in src/nasa_mouse_glare/README.md.

The expiMap/scArches handoff and architecture notes are documented in src/expiMap_scarches/EXPIMAP_HANDOFF.md.

Current expiMap run summaries and preprocessing comparisons are documented in docs/expimap_results.md.

The liver query-extension de novo-program analysis is documented in docs/expimap_de_novo_liver.md.

The tutorial-style liver expiMap run with HVG filtering, a deeper reference model, and HSIC de novo query nodes is documented in docs/expimap_tutorial_style_liver.md.

Accession-aware direct-model validation and the larger ARCHS4 reference seed-stability result are documented in docs/expimap_accession_validation.md and docs/expimap_reference_seed_stability.md. Condition-specific GC/FLT expiMap clustering is documented in docs/expimap_condition_clustering.md.

The skeletal-muscle pathway prior-work check is documented in docs/expimap_skeletal_muscle_prior_work.md.

The OntoVAE parallel pipeline, outputs, limitations, and expiMap comparison are documented in docs/ontovae_pipeline.md. The focused OntoVAE follow-up report with stable pathways, decoder genes, and plot links is documented in docs/ontovae_followup_report.md.

For NASA OSDR programmatic data access, see docs/osdr_api.md.