---
license: cc-by-nc-4.0
task_categories:
- tabular-classification
- question-answering
language:
- en
tags:
- spaceflight
- multi-omics
- biomedical
- benchmark
- LLM-evaluation
- epigenomics
- transcriptomics
- proteomics
- metabolomics
- microbiome
pretty_name: SpaceOmicsBench v3
size_categories:
- 10K<n<100K
---

# SpaceOmicsBench v3

**A Multi-Omics AI Benchmark for Spaceflight Biomedical Data**

SpaceOmicsBench v3 provides standardized ML and LLM evaluation infrastructure for spaceflight biomedical data from 4 human spaceflight missions (NASA Twins Study, Inspiration4, JAXA cfRNA, Axiom-2).

## Dataset Structure

### ML Track (Track A)
- `tasks/track_a/` — Task definitions (J1: phase classification, J2: clock acceleration)
- `tasks/track_c/` — Feature-level task definitions (C1: cfRNA DEG, C3: PPI benchmark)
- `data/tasks/` — Training data for J1 and J2 (n=20 samples)
- `data/axiom2_epigenetic/` — AX-2 epigenetic clock measurements (83 algorithms, 20 samples × 92 features)
- `data/external/string_ppi/` — STRING v12 PPI network topology features (17,995 genes × 4 features)
- `results/unified_baseline_results.json` — Complete leaderboard (26 tasks, 16+ baseline methods)
- `results/track_a/` — Stage-level results (TabPFN, ESM2, GNN, fusion, feature selection)

### LLM Track (Track B)
- `tasks/track_b/` — Question bank by category (12 categories, 270 questions)
- `evaluation/llm/question_bank_v3.json` — Full 270Q bank with ground truth, difficulty, uncertainty flags
- `evaluation/llm/annotation_schema.json` — Scoring rubric and 5-dimension framework

## Key Statistics
- **26 ML tasks** across 9 omics modalities (19 from v2 + 7 new)
- **270 LLM evaluation questions** (12 categories, 4 difficulty levels)
- **4 spaceflight missions**: NASA Twins (340d), Inspiration4 (3d), JAXA cfRNA, Axiom-2
- **Crew IDs**: Anonymized (C001–C004 for I4, A001–A004 for AX-2)

## Citation

If you use this dataset, please cite:

```
Jang, J. (2026). SpaceOmicsBench: A Multi-Omics AI Benchmark for Spaceflight Biomedical Data (v3).
Submitted to NeurIPS 2026 Datasets & Benchmarks Track.
GitHub: https://github.com/jang1563/SpaceOmicsBench
```

## License
- **Data**: CC BY-NC 4.0
- **Code**: MIT License

## Contact
JangKeun Kim (jak4013@med.cornell.edu) — Weill Cornell Medicine