Re-analysis of stimulated vs resting Effector CD4⁺ T cells from the human immune cell atlas (Calderon et al.; ATAC GSE118189), using GEO raw count matrices and methods aligned to the paper (TMM, voom, limma, BH q < 0.01, |log₂FC| > 1).
| Item | Description |
|---|---|
01_matrices_metadata.Rmd |
Inspect ATAC/RNA inputs, build sample_metadata.txt, export Effector CD4⁺ sample IDs (~1 min) |
02_effector_cd4_limma_PCA_integration.Rmd |
PCA, differential ATAC/RNA, integration plots, comparison to paper supplementary tables (~8 min) |
01_matrices_metadata.html / 02_effector_cd4_limma_PCA_integration.html |
Pre-rendered reports (open in a browser) |
results/ |
PCA, volcano, Venn, and integration figures (PDF/PNG) |
sample_metadata.txt |
Sample × cell type × condition × assay flags |
Supplementary_data_3_ATAC_stimulation_DA_peaks.txt |
Paper ATAC DA table (~40 MB; peak-level logFC comparison) |
Supplementary_data_4_RNA_stimulation_DE_genes.txt |
Paper RNA DE table (gene-level logFC comparison) |
GEO count matrices and Supplementary_tables.xlsx are not committed: see data/README.md.
# 1. Clone and enter the repo
git clone https://github.com/emrunali/atac-rna-effector-cd4.git
cd atac-rna-effector-cd4
# 2. Download GEO / supplementary files into data/ (see data/README.md)
# 3. Install R packages (Bioconductor): limma, edgeR, ChIPseeker, TxDb.Hsapiens.UCSC.hg19.knownGene, org.Hs.eg.db, ...
# 4. Render notebooks (watch chunk progress with quiet = FALSE)
Rscript -e 'rmarkdown::render("01_matrices_metadata.Rmd", quiet = FALSE)'
Rscript -e 'rmarkdown::render("02_effector_cd4_limma_PCA_integration.Rmd", quiet = FALSE)'Requires Pandoc for HTML (brew install pandoc on macOS).
flowchart LR
A[GEO ATAC counts] --> C[Atlas CPM filter]
B[GEO RNA raw counts] --> D[CD4+ CPM filter]
C --> E[Effector CD4+ subset]
D --> E
E --> F[limma-voom + duplicateCorrelation]
F --> G[Volcanos / PCA / Venn / scatter]
F --> H[Compare to paper Supp. Data 3–4]
| Step | Paper | This repository |
|---|---|---|
| RNA quantification | Kallisto + tximport estimated counts (Gencode v25) | NCBI raw counts from GSE118165 (GSE118165_raw_counts_GRCh38.p13_NCBI.tsv.gz) |
| Gene filter | Protein-coding, then CPM > 10 in ≥2 replicates → 13,512 genes tested | CPM > 10 on CD4⁺ columns first, then protein-coding via Entrez → 7,644 genes tested |
| ATAC filter | CPM ≥ 1 in ≥2 samples on full atlas → 671,448 peaks | Same rule → 734,428 peaks tested (GEO peak set) |
| Normalization | TMM | TMM |
| Testing | voom + limma | voom + limma |
| RNA design | Donor in design matrix | 6 libraries → donor via duplicateCorrelation(block = donor) + ~ condition |
| ATAC design | Donor + TSS enrichment | ~ tss_z + condition + duplicateCorrelation(block = donor) |
| Significance | BH q < 0.01, |log₂FC| > 1 | Same thresholds |
- Sample depth — Only 6 Effector CD4⁺ libraries are present in both GEO matrices (donors 1001, 1002, 1003, 1004; 1002 lacks resting, 1004 lacks stimulated). The paper reports up to four donors with balanced resting/stimulated pairs per cell type.
- RNA matrix — Paper tximport counts on Ensembl IDs vs GEO NCBI Entrez raw counts → fewer genes pass CPM > 10 (7,644 vs 13,512).
- Filter order (RNA) — Paper: protein-coding then CPM. Here: CPM then protein-coding (same as internal atlas notebooks; modest effect on gene count).
- Peak / gene universe — Slightly different consensus peak and gene catalogs on GEO vs in-house processing.
Despite different numbers of significant features, effect sizes agree strongly with the publication tables where features overlap.
| Metric | This analysis | Calderon et al. (reported) |
|---|---|---|
| RNA libraries | 6 | ~8 (4 donors × 2 conditions) |
| Genes tested | 7,644 | 13,512 |
| DEGs (q<0.01, |log₂FC|>1) | 170 up / 105 down (275 total) | 584 up / 282 down |
| Peaks tested | 734,428 | 671,448 |
| DARs (q<0.01, |log₂FC|>1) | 18,707 up / 2,047 down (20,754 total) | 20,210 up / 2,642 down |
| Pearson r (log₂FC vs paper, overlapping features) | RNA 0.974 (ENSEMBL join) | — |
| ATAC 0.988 (peak_id join) | — | |
| Proximal DAR–DEG overlap (±50 kb TSS) | 129 genes | — |
Pre-rendered outputs: results/volcano_RNA_effector_CD4.pdf, results/pca_*.pdf, results/venn_DAR_DEG_effector_CD4.png, results/scatter_ATAC_RNA_effector_CD4.pdf (ATAC volcano is in the HTML report; full-resolution PDF is gitignored due to size).
Volcano labels and supplementary-table ranks recover the expected T cell activation program, consistent with Calderon et al.:
- Immediate-early transcription factors: FOS, FOSB, JUN family (RNA and/or ATAC-linked peaks).
- Cytokine / effector genes: IL2, IL21, and related stimulation-induced transcripts (RNA).
- Cytotoxic / effector-associated genes seen in activated T profiles (e.g. GZMB in comparable analyses).
- Th17 / inflammatory axis: IL17F (and related stimulation-induced genes in this subset).
- Coordinated chromatin–expression changes: among genes significant in both assays at proximal promoters/enhancers, the majority show open & up (78 genes) or open & down / closed & down patterns matching directional accessibility–expression coupling in the paper’s framework (13 closed & down, 36 open & down, 2 closed & up).
Directional overlap (proximal DAR gene × DEG, n = 129):
| Pattern | Count |
|---|---|
| Open & RNA up | 78 |
| Open & RNA down | 36 |
| Closed & RNA down | 13 |
| Closed & RNA up | 2 |
.
├── README.md
├── 01_matrices_metadata.Rmd / .html
├── 02_effector_cd4_limma_PCA_integration.Rmd / .html
├── sample_metadata.txt
├── exported/effector_cd4_sample_ids.txt
├── data/
│ ├── README.md
│ └── GSE118165_GSM_to_sample_id.tsv
├── Supplementary_data_3_ATAC_stimulation_DA_peaks.txt
├── Supplementary_data_4_RNA_stimulation_DE_genes.txt
└── results/
- Calderon D, Nguyen MLT, Mezger A, Kathiria A, Müller F, Nguyen V, Lescano N, Wu B, Trombetta J, Ribado JV, Knowles DA, Gao Z, Blaeschke F, Parent AV, Burt TD, Anderson MS, Criswell LA, Greenleaf WJ, Marson A, Pritchard JK. Landscape of stimulation-responsive chromatin across diverse human immune cells. Nat Genet. 2019 Oct;51(10):1494-1505. doi: 10.1038/s41588-019-0505-9. Epub 2019 Sep 30. PMID: 31570894; PMCID: PMC6858557.
Note: If you use this code, please cite the original atlas publication and GEO accessions GSE118165 (RNA) and GSE118189 (ATAC).
MIT — see LICENSE.