# Human Heart Sex-Biased Chromatin & Gene Expression Analysis ## Overview This project investigates sex differences in chromatin accessibility (snATAC-seq) and gene expression (bulk RNA-seq) in human left ventricle (LV) heart tissue. Analyses include: - Differentially Accessible Regions (DAR): female vs. male (snATAC-seq, snATAC CM peaks) - Differentially Expressed Genes (DEG): old vs. young, stratified by sex (GTEx bulk RNA-seq LV) - Hormone receptor peak annotation: ER (estrogen receptor) and AR (androgen receptor) overlap with DARs --- ## Data Sources ### snATAC-seq (chromatin accessibility) - **Dataset**: GSE156704 — adult human heart snATAC-seq - **Samples**: 25 total — 12 female, 13 male (paired-end replicates per individual) - **Cell type**: Cardiomyocytes (CM) - **Peak set**: 158,489 merged peaks across all samples - **Count matrix**: `analysis/sn_ATAC_analysis/count_matrix_snPeak.txt` - Format: tab-separated, columns = chr / start / end / sample_S1 / sample_S2 ... - Equivalent file at: `ATAC_heart_openbed/counts_test/count_matrix.txt` (long column names) - Also available at: `/home/zhangbo/temp/Human_heart_GSE156704/merged_Audlt_snATAC_peak_counts.txt` (no `end` column) ### Bulk RNA-seq (gene expression) - **Dataset**: GTEx v8 — Heart Left Ventricle - **Samples**: 187 male (sex1), 91 female (sex2) - **Metadata**: `analysis/meta_sex1.txt`, `analysis/meta_sex2.txt` - Columns: sample_id / sex / age_band / age_group (young: 20–39, old: 40+) - **Count tables**: `getx_LV/` --- ## Directory Structure ``` heart/ ├── analysis/ │ ├── DAR_analysis_v6.R # DAR: female vs male, k=1~6 sweep, snATAC (bulk-merged peaks) │ ├── DAR_analysis_v7.R # DAR: female vs male, k=6 only (reference pipeline style) │ ├── DEG_sex1.R # DEG: old vs young, male (edgeR, v1) │ ├── DEG_sex2.R # DEG: old vs young, female (edgeR, v1) │ ├── DEG_sex1_v2.R # DEG: old vs young, male (DESeq2, v2) │ ├── DEG_sex2_v2.R # DEG: old vs young, female (DESeq2, v2) │ ├── meta_sex1.txt # GTEx male sample metadata │ ├── meta_sex2.txt # GTEx female sample metadata │ ├── sn_ATAC_analysis/ │ │ ├── count_matrix_snPeak.txt │ │ ├── heart_DAR_female_vs_male_RUVr_k{1-6}.csv # All peaks, per k │ │ ├── heart_DAR_female_vs_male_RUVr_k{1-6}_filter.csv # Significant DARs, per k │ │ ├── heart_DAR_female_vs_male_RUVr_k{1-6}_MORE.csv # Female > male, per k │ │ ├── heart_DAR_female_vs_male_RUVr_k{1-6}_LESS.csv # Female < male, per k │ │ ├── heart_DAR_female_vs_male_RUVr_k{1-6}_normCPM.csv │ │ ├── heart_DAR_female_vs_male_k_comparison_summary.csv # k=1~6 DAR count table │ │ ├── heart_DAR_overlap_all/MORE/LESS_k.csv # 6×6 pairwise overlap matrices │ │ ├── heart_DAR_peak_presence_across_k.csv │ │ └── results_k6/ # v7 reference pipeline output (k=6) │ │ ├── DAR_heart_femaleVSmale_all.csv │ │ ├── Sig_DAR_heart_femaleVSmale_FDR001_FC0585.csv │ │ ├── Sig_DAR_heart_femaleVSmale_FDR001_FC0585_UP.csv # female > male │ │ ├── Sig_DAR_heart_femaleVSmale_FDR001_FC0585_Down.csv # female < male │ │ ├── Normalized_cpm_RUV6.csv │ │ └── overlap_Yu_k6_identical_peaks.csv # Peaks shared with Yu_result │ └── DEG_LV/ │ ├── male/ # v1 edgeR DEG results (sex1) │ └── female/ # v1 edgeR DEG results (sex2) │ ├── snACAT_CM/ # snATAC CM-specific analysis (separate cohort) │ ├── counts/count_matrix_snACAT_CM.txt │ └── DAR_results/ │ ├── snACAT_CM_DAR_female_vs_male_RUVr_k6.txt │ ├── snACAT_CM_DAR_female_vs_male_RUVr_k6_filter.txt │ └── snACAT_CM_RUVr_k{1-6}_normCounts/PCA/RLE files │ ├── hormon/sn_atac/ # Hormone receptor annotation of snATAC peaks │ ├── snATAC_CM_peak_DAR_ER_AR.bed # All 158,489 peaks with DAR/ER/AR flags (TRUE/FALSE) │ ├── DAR_AR.bed # DAR=TRUE & AR=TRUE (503 peaks) │ ├── DAR_ER.bed # DAR=TRUE & ER=TRUE (1,069 peaks) │ ├── OCR_AR.bed # AR=TRUE, any peak (11,083 peaks) │ └── OCR_ER.bed # ER=TRUE, any peak (22,193 peaks) │ ├── Yu_result/CM_r_output/ # External collaborator DAR results (Yu lab, CM) │ ├── DAR_CM_femaleVSmale_ruvk6_exactTest.csv │ └── DAR_CM_femaleVSmale_ruvk6_exactTest_filtered.csv # 10,572 DARs │ ├── amina_res/ # External collaborator DAR results (Amina, k=4) │ └── DAR_CM/FB_*_FDR001_logFC58.csv/.bed │ └── adt_heart_narrowPeak/ # Individual sample MACS2 peak calls └── peak_union/ # Merged/union peak sets ``` --- ## Analysis Methods ### DAR Analysis (snATAC-seq) **Script**: `analysis/DAR_analysis_v6.R` (k sweep) / `DAR_analysis_v7.R` (k=6) Pipeline: 1. Load count matrix (158,489 peaks × 25 samples) 2. Build metadata: 12 female, 13 male 3. **Filter**: CPM > 1 in ≥ 12 samples (`min(table(sex))`) 4. TMM normalization (`calcNormFactors`) 5. Pre-batch PCA (log1p + prcomp) 6. RUVr batch correction: - edgeR GLM fit (common + tagwise dispersion) - Deviance residuals → `RUVr(seqUQ, k=k, residuals)` 7. DAR: edgeR `exactTest(pair=c("male","female"))` - logFC > 0 = female > male (MORE) - logFC < 0 = female < male (LESS) 8. Cutoffs: FDR < 0.001, |logFC| > log2(1.5) ≈ 0.585 9. Per-k: RLE plot, PCA, volcano plot, normalized CPM export 10. Overlap matrices across k values (6×6) **k=1~6 DAR count summary**: | k | MORE (F>M) | LESS (F 1 in ≥ min(group size) samples 4. RUVr batch correction (k=1~6) 5. DEG: edgeR exactTest (old vs young) [v1] or DESeq2 Wald test [v2] 6. Cutoffs: FDR < 0.05 (v1) / FDR < 0.01 (v2), |logFC| > 0.58 7. Per-k: volcano plots, UP/DOWN gene tables 8. Overlap matrices across k values (6×6, UP and DOWN separate) ### Hormone Receptor Annotation **Tool**: bedtools intersect Annotation file: `hormon/sn_atac/snATAC_CM_peak_DAR_ER_AR.bed` - Column 4 (DAR): overlaps with significant DAR peaks (female vs male, k=6) - Column 5 (ER): overlaps with ER ChIP-seq peaks (`ER_peaks.bed`) - Column 6 (AR): overlaps with AR binding sites (`masterlist_DHSs_myResults_nonovl_any_chunkIDs.bed`) | Annotation | Peaks | |-----------|-------| | DAR | 6,452 | | ER | 22,193 | | AR | 11,083 | | ER & AR | 1,909 | | DAR & ER | 1,069 | | DAR & AR | 503 | ### Cross-study Peak Overlap (Yu lab vs this study) **Files compared**: - Yu lab: `Yu_result/CM_r_output/DAR_CM_femaleVSmale_ruvk6_exactTest_filtered.csv` - 10,572 DARs; snATAC CM-specific cohort (5F / 5M); logFC > 0 = female > male - Yu MORE (F>M): 5,898 | Yu LESS (F male) - Yu LESS & mine LESS: 1,463 peaks (both female < male) - Discordant (opposite): 1,307 peaks **Peaks significant in both studies** (FDR < 0.001 & |logFC| > 0.585 in both): - 208 peaks — saved to `analysis/sn_ATAC_analysis/results_k6/overlap_Yu_k6_identical_peaks.csv` - Direction concordance: 114/208 (54%); 17 MORE/MORE, 97 LESS/LESS, 81 Yu-MORE/Mine-LESS, 13 Yu-LESS/Mine-MORE - logFC Pearson r = 0.31 (mean |logFC|: Yu = 1.08, mine = 0.83) - Overall logFC correlation (all 10,533 matched peaks): r = −0.02 **Interpretation**: Low replication (22% at FDR < 0.05) and near-random direction concordance (~42%) indicate the two datasets are largely independent. Key differences: different patient cohorts, different sample sizes (10 vs 25), and potentially different cell-type compositions (CM-specific vs bulk snATAC). The 208 peaks significant in both likely represent the most robust sex-biased chromatin loci in heart CM, predominantly with LESS direction (female < male accessibility). --- ## Key File Formats - **DAR result CSVs**: row = peak ID (`chr:start-end`), columns = genes / logFC / logCPM / PValue / FDR [/ CPM per sample] - **BED files** (4-column): chr / start / end / peak_ID (`chr_start_end`) - **Count matrices**: tab-separated, header row, first 3 cols = chr / start / end (or chr / start) - **Annotation BED**: chr / start / end / DAR / ER / AR (TRUE/FALSE) --- ## Notes - snATAC peak IDs use two formats across files: - Colon-dash: `chr1:9955-10505` (R CSVs) - Underscore: `chr1_9955_10505` (Yu lab, BED 4th column) - v6 and v7 DAR scripts are equivalent at k=6 after CPM filter alignment; v6 runs k=1~6 sweep, v7 runs k=6 only following the original reference pipeline structure - GTEx sex labels: sex1 = male, sex2 = female