README.md

---
dataset_name: riken2018_pbmc_supercentenarians
annotations_creators:
  - expert-generated
language:
  - en
multilinguality: "no"
pretty_name: PBMCs from Supercentenarians and Controls (Hashimoto et al. 2018)
task_categories:
  - cell-type-classification
size_categories:
  - 100K<n<1M
license: cc-by-4.0
dataset_type: biomedical
---

# Dataset Card for `riken2018_processed`

## Dataset Summary

This dataset is derived from single-cell RNA-seq of peripheral blood mononuclear cells (PBMCs) from supercentenarians (ages 110+) and younger controls, processed from the study:

> *Single-cell transcriptomics reveals expansion of cytotoxic CD4 T cells in supercentenarians*  
> — Hashimoto et al., *PNAS* (2019)  
> [DOI:10.1073/pnas.1907883116](https://doi.org/10.1073/pnas.1907883116)

The final object includes cells from three experimental batches (firstrun, SC1, SC2) and supports aging-focused immunology research.

## Transformation Summary

Raw data was downloaded from [http://gerg.gsc.riken.jp/SC2018/](http://gerg.gsc.riken.jp/SC2018/) and processed with the following steps:

1. **Download and Extraction**:
   - Retrieved UMI matrix (`01.UMI.txt.gz`) and cell barcodes (`03.Cell.Barcodes.txt.gz`).
   - Extracted expression matrices from `SC1.tar` and `SC2.tar` using `scanpy.read_10x_mtx`.

2. **UMI Matrix Assembly**:
   - Constructed an `AnnData` object from the UMI count matrix and matched barcodes.
   - Gene names were mapped to Ensembl IDs using SC1 reference.

3. **Batch Merging**:
   - Merged `firstrun`, `SC1`, and `SC2` into one `AnnData` object using `scanpy.concat`, with batch labels preserved.

4. **Metadata Curation**:
   - Filled in missing columns for sample origin (`SC1`, `SC2`, etc.) based on batch labels.
   - Added standardized columns: `age_int`, `assay_simple`, `cell_type`, and `centenarian_status`.

5. **Output**:
   - Saved final object to `raw/riken2018/processed/riken2018_processed.h5ad`.

## Supported Tasks and Benchmarks

- **Immune Aging Analysis**: Especially suited for studying the immune profile of extreme aging.
- **CD4 T Cell Phenotyping**: Detects rare CD4 cytotoxic T cell expansions.
- **Batch Integration**: Includes multiple experimental runs merged with consistent annotations.

## Languages

Textual metadata and annotations are in English.

## Dataset Structure

### Data Instances

Each instance represents a single PBMC with:

- Raw UMI expression data
- Batch origin (firstrun, SC1, SC2)
- Age group metadata (`centenarian_status`, `age_int`)
- Cell type label (`PBMC`)

### Data Splits

- No formal train/test split provided. Users may stratify by `centenarian_status`.

## Dataset Creation

### Curation Rationale

The dataset was assembled to study cellular aging phenotypes by comparing PBMC populations between centenarians and controls, with a focus on adaptive immunity.

### Source Data

- 7 supercentenarians and 5 controls
- Collected and sequenced using 10x Genomics Chromium 3' technology
- Published by RIKEN Center for Integrative Medical Sciences

### Preprocessing Details

- Raw count matrix: `01.UMI.txt.gz`
- Barcode matching and gene name alignment using 10x `SC1` reference
- Minor cleanup of missing metadata and consistent column naming
- Concatenation with `scanpy` after aligning gene and barcode identifiers

## Licensing Information

This dataset is distributed under the **Creative Commons BY 4.0** license. Please cite the original paper when using this dataset in publications.

## Citation

```bibtex
@article{hashimoto2019supercentenarians,
  title={Single-cell transcriptomics reveals expansion of cytotoxic CD4 T cells in supercentenarians},
  author={Hashimoto, Kosuke and Kouno, Tsukasa and Ikawa, Tomokatsu and et al.},
  journal={Proceedings of the National Academy of Sciences},
  volume={116},
  number={48},
  pages={24242--24251},
  year={2019},
  publisher={National Academy of Sciences}
}