Processed mission-held-out transcriptomics folds for evaluating whether machine-learning and foundation-model methods generalize spaceflight biological signatures across missions.

Public status: v7.1.2 public-card/metadata patch over canonical v7.1 results

Dataset freeze: 2026-03-01

Patch scope: documentation, public metadata, and access guidance. It does not introduce new benchmark result generation.

Code and full documentation: https://github.com/jang1563/GeneLab_benchmark

Maintainer / citation author: JangKeun Kim, Weill Cornell Medicine.

What Is In This Dataset

This Hugging Face dataset contains self-contained public fold packages from the GeneLab Benchmark v1-v7 surface. Each fold holds out one mission as the test set and provides all files needed to train on the remaining missions and evaluate on the held-out mission.

Public package item	Description
`train_X.csv`, `test_X.csv`	Sample-by-gene expression matrices
`train_y.csv`, `test_y.csv`	Binary labels: `1 = Flight`, `0 = Ground`
`train_meta.csv`, `test_meta.csv`	Sample-level metadata used for fold auditing
`fold_info.json`	Held-out mission, train missions, and sample-count audit metadata
`selected_genes.txt`	Fold-specific genes selected from training missions only
`task_info.json`	Task-level metadata and source summary

The web Dataset Viewer is disabled because these are high-dimensional sample-by-gene matrices plus JSON artifacts. Use direct downloads for reliable access.

Public Fold Layout

genelab-benchmark/
├── A2_gastrocnemius_lomo/
│   ├── task_info.json
│   ├── fold_RR-1_test/
│   ├── fold_RR-5_test/
│   └── fold_RR-9_test/
├── A4_thymus_lomo/
│   ├── task_info.json
│   ├── fold_MHU-1_test/
│   ├── fold_MHU-2_test/
│   ├── fold_RR-6_test/
│   ├── fold_RR-9_test/
│   └── fold_RR-23_holdout/
├── A5_skin_lomo/
│   ├── task_info.json
│   ├── fold_MHU-2_test/
│   ├── fold_RR-6_test/
│   ├── fold_RR-7_test/
│   └── fold_RR-7_holdout/
├── A6_eye_lomo/
│   ├── task_info.json
│   ├── fold_RR-1_test/
│   ├── fold_RR-3_test/
│   └── fold_OSD-397_test/
├── v4/evaluation/
├── v5/evaluation/
└── v6/evaluation/

fold_OSD-397_test is the stable public label for the third A6 eye fold.

Scope

Dimension	Coverage
Full v1-v7 benchmark surface	8 tissues
Public source catalog	24+ NASA OSDR accessions
Processed sample scope	600+ binary/control samples across release layers
v4 multi-method evaluation	8 tissues x 8 classifiers x 4 feature types = 256 evaluations
Public HF fold package	4 reviewer-facing LOMO tasks plus selected result artifacts

The full GitHub benchmark also includes historical v2-v7 extensions for temporal dynamics, cross-species analysis, single-cell and spatial pilots, foundation-model comparisons, graph/network baselines, and biological interpretation layers. This HF repository is optimized for processed dataset access; GitHub is the complete methods, code, and release-documentation surface.

Download Example

from huggingface_hub import hf_hub_download
import pandas as pd

repo_id = "jang1563/genelab-benchmark"
fold = "A5_skin_lomo/fold_RR-7_test"

def hf_csv(name):
    return pd.read_csv(
        hf_hub_download(
            repo_id=repo_id,
            filename=f"{fold}/{name}",
            repo_type="dataset",
        ),
        index_col=0,
    )

train_X = hf_csv("train_X.csv")
train_y = hf_csv("train_y.csv").iloc[:, 0]
test_X = hf_csv("test_X.csv")
test_y = hf_csv("test_y.csv").iloc[:, 0]
test_meta = hf_csv("test_meta.csv")

print(train_X.shape, train_y.shape, test_X.shape, test_y.shape)

Download a complete task:

from huggingface_hub import snapshot_download

snapshot_download(
    repo_id="jang1563/genelab-benchmark",
    repo_type="dataset",
    allow_patterns="A5_skin_lomo/**",
    local_dir="./data/genelab-benchmark",
)

File Contract

File	Contract
Feature matrices	Rows are sample IDs; columns are Ensembl mouse gene IDs
Expression values	Log2(DESeq2 size-factor normalized counts + 1)
Gene selection	Top 75th percentile variance, computed on training missions only
Labels	Binary Flight/Ground labels
Metadata	Sample and fold metadata for auditability

The fold design prevents test-mission leakage by applying variance filtering inside each training split.

Evaluation Summary

The canonical result source is docs/CANONICAL_RESULTS_V7_1.md in the GitHub repository.

Result surface	Takeaway
Multi-method benchmark	PCA-LR is the strongest 8-tissue gene-level baseline in v4, with mean AUROC 0.776.
Best tissue rows	Thymus 0.948, colon 0.921, lung 0.901, kidney 0.829 across best method-feature combinations.
Cross-mission transfer	Thymus and gastrocnemius show the strongest mission-transfer signal; liver and kidney are harder.
Pathway features	Pathway representations rescue some weaker gene-level tissues, especially kidney and eye.
Foundation models	Tested gene-expression foundation models underperform tuned classical baselines on small-n bulk RNA-seq mission shift.
Held-out validation	Thymus RR-23 AUROC 0.905; skin RR-7 AUROC 0.885.

Intended Use

Use this dataset to:

evaluate spaceflight transcriptomics classifiers under mission-held-out shift;
compare classical ML, foundation-model, and adapter methods on fixed folds;
test preprocessing or feature representations without changing test missions;
reproduce public benchmark summaries from the GitHub repository.

For full methods and release status, use the GitHub documentation and release manifest.

Release Labels

Surface	Public label
v7.1 GeneLab Benchmark	Canonical historical result surface and citation target
v7.1.2 public-card patch	Documentation and metadata patch over v7.1 results

This HF dataset card describes the v7.1 public fold package with the v7.1.2 public-card patch.

Citation

Please cite the software and benchmark using the GitHub CITATION.cff metadata.

@dataset{kim2026genelab,
  title = {SpaceBio-Bench / GeneLab Benchmark: Mission-Held-Out Spaceflight Transcriptomics Benchmark},
  author = {Kim, JangKeun},
  year = {2026},
  url = {https://huggingface.co/datasets/jang1563/genelab-benchmark},
  note = {v7.1.2 documentation, public-card, and metadata patch over canonical v7.1 results; data freeze 2026-03-01}
}

Source data: NASA Open Science Data Repository (OSDR), https://osdr.nasa.gov/bio/repo/.

License

Processed dataset package: CC-BY-4.0
Code: MIT, in the GitHub repository
Source data: NASA OSDR public data; follow individual source-dataset terms

Downloads last month: 320

Total file size:

400 MB