scShapeBench / README.md
btd2026's picture
Upload README.md with huggingface_hub
fcb69dc
|
raw
history blame
3.14 kB
metadata
license: other
tags:
  - single-cell
  - rna-seq
  - scRNA-seq
  - h5ad
  - anndata
  - genomics
  - benchmark
  - shape-analysis
size_categories:
  - 1M<n<10M
task_categories:
  - tabular-classification
  - tabular-regression
pretty_name: scShapeBench

scShapeBench

A benchmark dataset for single-cell shape analysis, containing 2,547,517 cells across 105 real-world single-cell RNA-seq datasets in AnnData (.h5ad) format.

Dataset Summary

scShapeBench is a curated collection of real-world single-cell gene expression datasets assembled for benchmarking computational methods in single-cell shape analysis. Each dataset is stored as an individual AnnData file with precomputed PCA embeddings and Leiden clustering.

  • Total cells: 2,547,517
  • Total datasets: 105
  • Total size: ~116 GB
  • Format: AnnData (.h5ad)
  • Data type: Real-world (synthetic data to be added in a future release)

Dataset Structure

scShapeBench/
├── data/
│   ├── SCD-0001.h5ad
│   ├── SCD-0002.h5ad
│   ├── ...
│   └── SCD-0112.h5ad
├── cell_metadata.csv    # Combined cell-level metadata (2.5M rows)
├── gene_metadata.csv    # Combined gene-level metadata
├── dataset_index.csv    # Per-file summary: dimensions, size
├── croissant.json       # Croissant 1.1 metadata
└── README.md

File Naming

Files are named SCD-XXXX.h5ad where XXXX is a zero-padded index. The numbering is not contiguous (e.g., SCD-0031, SCD-0032, SCD-0034–0036 are absent).

Per-File Contents

Each .h5ad file contains:

Component Description
X Gene expression matrix (cells × genes), log-normalized
obs Cell metadata: n_genes, leiden (cluster assignment)
var Gene metadata: gene_ids, feature_types, genome, n_cells, highly_variable, etc.
obsm['X_pca'] Precomputed PCA embeddings
uns Clustering and HVG parameters

Dataset Index

The dataset_index.csv file provides per-file summary statistics:

Column Description
file_id Dataset identifier (e.g., SCD-0001)
filename Filename
n_cells Number of cells
n_genes Number of genes
file_size_bytes File size in bytes

Dataset sizes range from 1,163 cells (SCD-0006) to 434,340 cells (SCD-0037).

Usage

import scanpy as sc
import pandas as pd

# Load a single dataset
adata = sc.read_h5ad("data/SCD-0001.h5ad")
print(adata)

# Browse available datasets
index = pd.read_csv("dataset_index.csv")
print(index.sort_values("n_cells", ascending=False).head())

# Load cell metadata across all datasets
cell_meta = pd.read_csv("cell_metadata.csv")
print(cell_meta.groupby("file_id").size())

Croissant Metadata

This dataset includes a croissant.json file conforming to the Croissant 1.1 metadata standard. This enables programmatic discovery and loading of dataset metadata through compatible tools.

Citation

License

License to be determined.