Update README.md

README.md

@@ -7,75 +7,84 @@ language:
 multilinguality: "no"
 pretty_name: Single-cell Atlas of Human Blood Aging (Terekhova et al. 2023)
 task_categories:
-  - other
+  - cell-type-classification
 size_categories:
   - 1M<n<10M
 license: cc-by-4.0
 dataset_type: biomedical
 ---
 
-
 # Dataset Card for `tereshkova2023_processed`
 
 ## Dataset Summary
 
-This dataset is a single-cell multi-omic atlas of peripheral blood mononuclear cells (PBMCs) from 166 healthy human donors aged 25–85, comprising ~2 million single cells profiled using scRNA-seq, single-cell TCR/BCR-seq, and surface protein barcoding (CITE-seq). The dataset is a core resource from the 2023 Immunity paper:
+This dataset is a single-cell multi-omic atlas of peripheral blood mononuclear cells (PBMCs) from 166 healthy donors aged 25–85, comprising ~2 million cells profiled with scRNA-seq, TCR/BCR-seq, and surface protein barcoding (CITE-seq). It supports cell type annotation, immune aging trajectory modeling, and multimodal integration. It is described in:
 
 > *Single-cell atlas of healthy human blood unveils age-related loss of NKG2C+GZMB–CD8+ memory T cells and accumulation of type 2 memory T cells*  
 > — Terekhova et al., *Immunity*, Volume 56, Issue 12 (2023)
 
+## Transformation Summary
+
+The raw data was processed with the following steps:
+
+1. **File Acquisition**: Downloaded `.tar.gz` archives from Synapse (`all_pbmcs.tar.gz`, `raw_counts_h5ad.tar.gz`) and moved them into a structured `raw/terekhova2022` directory.
+2. **Extraction**: Extracted `pbmc_gex_raw_with_var_obs.h5ad` from the tarball.
+3. **Demultiplexing Assignments**: Downloaded and merged cell-level assignments from multiple `*_assignment_cell_patient.csv` files into a single table, then joined this with the `.obs` table in the AnnData object.
+4. **Metadata Integration**: Manually downloaded `mmc2.xlsx`, reshaped it to long format per tube/visit, and mapped donor metadata (age, sex, ethnicity, BMI, visit) to cells using the patient ID.
+5. **Annotation Cleanup**: Renamed columns for clarity (`nCount_RNA → raw_sum`, `Cluster_names → unique_cell_type`, etc.), dropped unused donor ID.
+6. **Final Output**: Saved the processed AnnData object as `processed/terekhova2022_processed.h5ad`.
+
 ## Supported Tasks and Benchmarks
 
-- **Immune Cell Type Annotation**: 55 PBMC subpopulations annotated across T cells, B cells, NK cells, myeloid, and progenitor compartments.
-- **Aging Trajectory Modeling**: Captures cell composition and transcriptional changes across a 60-year adult lifespan.
-- **TCR/BCR Clonality Analysis**: Includes paired VDJ sequencing enabling repertoire studies.
-- **Surface Protein and Transcriptome Integration**: Enables multi-modal modeling and CITE-seq analysis.
-- **Subtype Discovery**: Features rare populations like NKG2C+GZMB–CD8+ memory T cells, declining with age.
+- **Immune Cell Type Annotation**: 55 annotated PBMC subtypes spanning all major lineages.
+- **Aging Trajectory Modeling**: Suitable for lifespan analyses (ages 25–85).
+- **VDJ Repertoire Analysis**: Includes matched TCR/BCR sequences.
+- **Multimodal Learning**: Combines surface proteins and gene expression for CITE-seq tasks.
+- **Rare Subtype Discovery**: Detects rare populations, such as NKG2C+GZMB–CD8+ memory T cells.
 
 ## Languages
 
-Textual metadata (e.g. donor annotations) are in English.
+Textual metadata (e.g. donor demographics) are in English.
 
 ## Dataset Structure
 
 ### Data Instances
 
-Each data instance corresponds to a single cell with the following modalities:
+Each instance corresponds to a single cell, with:
 
-- Gene expression (UMI counts)
-- Feature barcoding (surface protein levels)
-- TCR / BCR sequences (if available)
-- Donor-level metadata (age, sex, visit ID, etc.)
+- Gene expression (UMIs)
+- Surface protein (CITE-seq)
+- VDJ sequences (where available)
+- Annotated metadata (age, sex, visit, ethnicity, etc.)
 
 ### Data Splits
 
-The dataset includes 317 samples from 166 individuals, some sampled longitudinally. Split recommendations:
-
-- `train`: Samples from age groups 25–64
-- `test`: Samples from 65+
+- `train`: Age groups 25–64
+- `test`: Age 65+
 
 ## Dataset Creation
 
 ### Curation Rationale
 
-Aging affects immune composition and function. This dataset was designed to provide a high-resolution reference of healthy immune aging, addressing limitations of smaller cohort studies.
+Designed to enable high-resolution studies of immune aging using a large, high-quality cohort of healthy adults, overcoming prior limitations in sample size and longitudinal tracking.
 
 ### Source Data
 
-- PBMCs collected under IRB-approved protocol at Washington University in St. Louis.
-- Individuals were fasting, non-obese, healthy adults with no chronic inflammatory conditions.
-- 10x Genomics 5’ v2 with TotalSeq-C feature barcoding and VDJ enrichment.
+- IRB-approved PBMC collection at Washington University in St. Louis.
+- Subjects: healthy, fasting, non-obese, free of chronic inflammatory diseases.
+- Assays: 10x 5′v2 with TotalSeq-C and VDJ enrichment.
 
-### Preprocessing
+### Preprocessing Details
 
-- Demultiplexing with HTO (Seurat HTODemux) and genotype-based (Souporcell) methods.
-- Quality control: filtering based on gene count, mitochondrial content, and doublet detection.
-- Batch correction via Harmony.
-- Clustering and annotation via Seurat and Scanpy.
+- Demultiplexing: HTO (Seurat) + genotype-based (Souporcell)
+- Filtering: Gene count, mitochondrial percentage, doublets
+- Batch correction: Harmony
+- Clustering and annotation: Seurat + Scanpy
+- Metadata merge from Excel and CSV sources, structured by tube ID
 
 ## Licensing Information
 
-This dataset is distributed under **Creative Commons BY 4.0** license. Cite the original paper when using this dataset.
+This dataset is distributed under the **Creative Commons BY 4.0** license. Please cite the original paper in any publications or derived work.
 
 ## Citation