sc-human-blood-aging-atlas

Study Description

This dataset accompanies the study “Single-cell atlas of healthy human blood unveils age-related loss of NKG2C+GZMB− CD8+ memory T cells and accumulation of type 2 memory CD4+/CD8+ T cells”, which provides a comprehensive single-cell transcriptomic and proteomic landscape of peripheral blood from healthy individuals across the adult lifespan.

Using single-cell RNA sequencing (scRNA-seq) integrated with TCR/BCR profiling and protein feature barcoding (20 antibodies), the study profiled 317 peripheral blood mononuclear cell (PBMC) samples from 166 individuals aged 25–85 years. The resulting dataset contains ~2 million immune cells, annotated into 50+ immune cell subpopulations.

Key findings include:

• Decline with age of a novel subset of NKG2C⁺GZMB⁻ CD8⁺ central memory T cells
• Accumulation with age of GZMK⁺ CD8⁺ effector memory T cells and HLA-DR⁺ CD4⁺ memory T cells
• Increased IL-4 potential in CD4⁺ and CD8⁺ Type 2 memory T cells, indicating a shift toward a type 2 immune phenotype with age
• Clonality analysis across T and B cell subsets reveals age-related changes in repertoire

This dataset serves as a rich resource for exploring age-associated immune remodeling and supports downstream machine learning and computational modeling efforts in aging, immunology, and single-cell systems biology.

---

Dataset Overview

This dataset contains multi-modal single-cell measurements from peripheral blood mononuclear cells (PBMCs) collected from 166 healthy individuals aged 25–85, spanning a 3-year period.

Attribute	Description
Individuals	166 healthy donors
Samples	317 PBMC samples (some donors sampled longitudinally)
Cells	~2 million high-quality single cells
Modalities	scRNA-seq, TCR-seq, BCR-seq, 20-plex protein barcoding
Format	.parquet (processed tables); .h5ad (raw gene counts & annotations)
Organism	Homo sapiens (human)
Tissues	Peripheral blood
Age Range	25 to 85 years old

---

Cell Type Annotations

The dataset includes high-confidence cell type annotations for ~2 million PBMCs, derived through integrated analysis of gene expression, surface protein markers, and immune receptor (TCR/BCR) features.

Key Details:

• Total subpopulations: Over 50 distinct immune cell types
• Annotation method:
  • Canonical marker genes + antibody-derived tags
  • Unsupervised clustering (Leiden/Louvain)
  • Expert-curated lineage assignments

Major Cell Classes:

• T cells: CD4⁺ naïve, central memory (Tcm), effector memory (Tem), regulatory (Treg), Th1/Th2/Th17 subsets, and CD8⁺ subtypes
• B cells: Naïve B, memory B, plasma cells
• NK cells: CD56⁺ bright and dim NK cells
• Monocytes/DCs: Classical, intermediate, non-classical monocytes; plasmacytoid/myeloid DCs
• Other: γδ T cells, MAIT cells, progenitor-like cells

Each .parquet file contains a cell_type_annotation column specifying the immune subpopulation, and a barcode column for cell tracking across modalities.

Note: The barcode column serves as the primary key for each single cell and can be used to map across modalities such as RNA expression, surface protein markers, and immune receptor features (TCR/BCR).

---

File Structure

This repository includes 14 .parquet files, each representing a unique PBMC sample from a specific donor/timepoint.

Each file includes:

• barcode: Unique cell ID
• cell_type_annotation: Assigned immune subpopulation
• Expression + surface protein features (e.g., nCount_RNA, CD4, CD8A, etc.)

These tables support:

• Cell type classification
• Dimensionality reduction (UMAP, t-SNE)
• Age-association modeling
• Multi-omic integration

Note: Raw .h5ad gene expression matrices are available via Synapse.

---

Quickstart (Python)

import pandas as pd
from huggingface_hub import hf_hub_download

# Example: Load one sample
file_path = hf_hub_download(repo_id="longevity-db/sc-human-blood-aging-atlas", filename="AS044.parquet")
df = pd.read_parquet(file_path)

# Preview
print(df.shape)
print(df[['barcode', 'cell_type_annotation']].head())

---

Suggested ML Use Cases

• 🧬 Cell type classification
• 🧓 Age prediction from immune profiles
• 🧭 Trajectory inference and pseudotime modeling
• 🧩 Dimensionality reduction + clustering
• 📊 Cell-type-specific differential expression with age
• 🔁 Transfer learning across donors or tissues

---

Related Resources

• 📄 Original Publication – Immunity, 2023
• 🧪 Synapse Project (Raw data)
• 🔬 CellxGene Viewer (if available) (optional)

---

Citation

Hafemeister et al., "Single-cell atlas of healthy human blood unveils age-related loss of NKG2C+GZMB− CD8+ memory T cells and accumulation of type 2 memory CD4+/CD8+ T cells."
Immunity (2023) — DOI: 10.1016/j.immuni.2023.10.003

Contributing

This dataset was created for the Longevity x AI Hackathon 2025. Contributions, improvements, and additional analyses are welcome! For questions or collaboration opportunities, please contact [your-email@domain.com].

This dataset represents a novel intersection of cancer genomics, digital pathology, and aging research, enabling unprecedented insights into the cellular basis of human aging.