Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,248 @@
+---
+language:
+- en
+tags:
+- dna-methylation
+- epigenetic-clock
+- promoter-methylation
+- global-methylation
+- breast-cancer
+- sub-saharan-africa
+license: cc-by-nc-4.0
+pretty_name: SSA Breast DNA Methylation Dataset (Women, Multi-ancestry)
+task_categories:
+- other
+size_categories:
+- 1K<n<10K
+---
+
+# SSA Breast DNA Methylation Dataset (Women, Multi-ancestry, Synthetic)
+
+## Dataset summary
+
+This dataset provides a **synthetic DNA methylation cohort** conceptually linked to women with invasive breast cancer across multiple ancestry groups, with emphasis on **Sub-Saharan Africa (SSA)** and comparable reference populations.
+
+It captures **high-level methylation features** rather than site-level arrays, grounded in key literature on epigenetic clocks, promoter CpG methylation of tumour suppressor genes, global hypomethylation, and race/ancestry-associated methylation patterns:
+
+- **Epigenetic age and age acceleration** (Horvath/Hannum-like clocks).
+- **Global DNA methylation** (e.g., LINE-1 / LUMA-like indices).
+- **Promoter methylation** of canonical tumour suppressors (BRCA1, RASSF1A, CDH1, CDKN2A/p16, DAPK1).
+- **Population-specific methylation signature clusters** informed by AA vs EA tumour methylation studies.
+
+All records are **fully synthetic**, derived from literature-based distributions and not real methylation arrays.
+
+
+## Cohort design
+
+### Sample size and populations
+
+- **Total N**: 10,000 synthetic DNA methylation profiles.
+- **Populations**:
+  - `SSA_West`: 2,000
+  - `SSA_East`: 2,000
+  - `SSA_Central`: 1,500
+  - `SSA_Southern`: 1,500
+  - `AAW` (African American women): 1,500
+  - `EUR` (European reference): 1,000
+  - `EAS` (East Asian reference): 500
+
+Demographics mirror other Electric Sheep Africa breast cancer modules:
+
+- Predominantly **women**, with ~1% male breast cancers.
+- Ages 18–90, with **younger SSA** vs **older EUR/EAS/AAW** distributions.
+
+
+## Epigenetic age and age-acceleration
+
+Variables:
+
+- `epigenetic_age` – DNA methylation age (years).
+- `age_accel_years` – epigenetic age minus chronological age.
+- `age_accel_category` – `Decelerated`, `On_track`, `Accelerated`.
+
+Anchoring:
+
+- **Horvath 2013** (multi-tissue clock) and follow-up reviews show DNAm age tracks chronological age with **median error ≈3–5 years**.
+- Age acceleration (Δage) typically lies within **±5–10 years**, with modest shifts by lifestyle, disease, and ancestry.
+
+In this dataset:
+
+- `age_accel_years` is drawn from category-specific distributions (e.g., Decelerated ≈ -5 ± 2; Accelerated ≈ +5 ± 3).
+- SSA and AAW populations have slightly higher proportions of `Accelerated` than EUR/EAS, reflecting literature suggesting modest differences in DNAm age acceleration.
+
+
+## Global DNA methylation
+
+Variables:
+
+- `global_methylation_category` – `Low`, `Intermediate`, `High`.
+- `global_methylation_beta` – continuous 0–1 index (LINE-1/LUMA-like surrogate).
+
+Anchoring:
+
+- Studies of **LINE-1** and luminometric methylation suggest **global hypomethylation** in roughly **25–40%** of breast cancer patients, associated with genomic instability and poor prognosis.
+
+Configuration:
+
+- SSA and AAW populations have somewhat higher `Low` global methylation fractions than EUR/EAS, consistent with elevated hypomethylation prevalence in some series.
+- `global_methylation_beta` is drawn around category means (e.g., Low ≈0.55, Intermediate ≈0.70, High ≈0.82) with small noise.
+
+
+## Promoter methylation of tumour suppressor genes
+
+Genes modeled:
+
+- `BRCA1`
+- `RASSF1A`
+- `CDH1` (E-cadherin)
+- `CDKN2A_p16`
+- `DAPK1`
+
+Variables per gene (example for BRCA1):
+
+- `BRCA1_promoter_status` – `Unmethylated`, `Partially_methylated`, `Hypermethylated`.
+- `BRCA1_promoter_beta` – approximate promoter CpG beta value (0–1).
+
+Anchoring:
+
+- Multiple tumour series report **promoter CpG island hypermethylation** of BRCA1, RASSF1A, CDH1, CDKN2A/p16, and DAPK1 in **~20–50%** of breast cancers, often enriched in **ER-negative, high-grade, and TNBC** tumours.
+- Studies from the Carolina Breast Cancer Study and other cohorts note **racial differences** in promoter methylation (e.g., some loci more frequently methylated in tumours from African American women).
+
+In this dataset:
+
+- Per-gene promoter status distributions are defined **by population**, with slightly higher `Hypermethylated` fractions for selected genes (e.g., BRCA1, RASSF1A) in `SSA_*` and `AAW` groups vs `EUR/EAS`.
+- Beta values are drawn from status-based distributions that approximate Illumina 450K/EPIC behaviour:
+  - `Unmethylated` ≈0.1±0.05.
+  - `Partially_methylated` ≈0.5±0.10.
+  - `Hypermethylated` ≈0.8±0.07.
+
+
+## Population-specific methylation signatures
+
+Variables:
+
+- `methylation_signature_cluster` – `Cluster_A_EUR_like`, `Cluster_B_AAW_like`, `Cluster_C_SSA_like`.
+
+Anchoring:
+
+- **Genome-wide 450K analysis** of tumours from African American vs European American women (e.g., Oncotarget 2013; Carolina Breast Cancer Study) reveals **distinct CpG methylation patterns**, including differences in global methylation and promoter methylation at key loci.
+
+In this dataset:
+
+- `Cluster_A_EUR_like` is predominant in `EUR/EAS` populations, but present at lower frequency in AAW/SSA.
+- `Cluster_B_AAW_like` is enriched in `AAW`, but also present in SSA and some EUR/EAS tumours.
+- `Cluster_C_SSA_like` is more frequent in SSA populations, capturing a stylized SSA-associated methylation signature.
+
+These clusters are **latent summary labels** rather than actual high-dimensional profiles, but they are consistent with reported patterns of ancestry-associated methylation.
+
+
+## File and schema
+
+### `dna_methylation_data.parquet` / `dna_methylation_data.csv`
+
+Each row represents a synthetic tumour DNA methylation profile with:
+
+- **Demographics**
+  - `sample_id`
+  - `population`
+  - `region`
+  - `is_SSA`
+  - `is_reference_panel`
+  - `sex`
+  - `age`
+
+- **Epigenetic age**
+  - `epigenetic_age`
+  - `age_accel_years`
+  - `age_accel_category`
+
+- **Global methylation**
+  - `global_methylation_category`
+  - `global_methylation_beta`
+
+- **Promoter methylation (per gene)**
+  - `{GENE}_promoter_status`
+  - `{GENE}_promoter_beta`
+
+- **Population signature**
+  - `methylation_signature_cluster`
+
+
+## Generation
+
+The dataset is generated using:
+
+- `dna_methylation/scripts/generate_dna_methylation.py`
+
+with configuration in:
+
+- `dna_methylation/configs/dna_methylation_config.yaml`
+
+and literature inventory in:
+
+- `dna_methylation/docs/LITERATURE_INVENTORY.csv`
+
+Generation steps:
+
+1. **Demographic cohort** – multi-ancestry cohort with age and sex distributions by population.
+2. **Age acceleration** – assign `age_accel_category` by population and draw `age_accel_years` from Horvath/Raj-informed distributions; compute `epigenetic_age`.
+3. **Global methylation** – assign `global_methylation_category` by population and draw `global_methylation_beta` from category-specific distributions.
+4. **Promoter methylation** – for each gene, assign qualitative status by population and sample a corresponding beta value.
+5. **Signature clusters** – assign `methylation_signature_cluster` by population, reflecting AA/EA and SSA/EUR patterns reported in genome-wide methylation studies.
+
+
+## Validation
+
+Validation is performed with:
+
+- `dna_methylation/scripts/validate_dna_methylation.py`
+
+and summarized in:
+
+- `dna_methylation/output/validation_report.md`
+
+Checks include:
+
+- **C01–C02** – Sample size and population counts vs config.
+- **C03** – Age acceleration category distributions by population.
+- **C04** – Global methylation category distributions by population.
+- **C05** – Promoter methylation status distributions by population and gene.
+- **C06** – Methylation signature cluster distributions by population.
+- **C07** – Missingness across all key columns.
+
+An absolute deviation tolerance of **10%** is used for categorical distributions, and the released dataset is configured to reach an overall validation status of **`PASS`**.
+
+
+## Intended use
+
+This dataset is intended for:
+
+- **Methods development** in integrating DNA methylation with other SSA synthetic modules (pathology, IHC, immune, systemic markers, microbiome).
+- **Simulation studies** on epigenetic age, promoter methylation, and ancestry-associated methylation signatures.
+- **Educational use** in teaching epigenetic concepts in cancer.
+
+It is **not intended** for:
+
+- Estimating exact methylation frequencies or beta-value distributions for any specific locus or ancestry.
+- Direct clinical interpretation or patient-level risk prediction.
+
+
+## Ethical considerations
+
+- All methylation profiles are simulated; no real 450K/EPIC or bisulfite assay data are used.
+- Ancestry-related differences are modeled to reflect literature trends but must not be interpreted as deterministic or stigmatizing.
+
+
+## License
+
+- License: **CC BY-NC 4.0**.
+- Free for non-commercial research, methods development, and education with attribution.
+
+
+## Citation
+
+If you use this dataset, please cite:
+
+> Electric Sheep Africa. "SSA Breast DNA Methylation Dataset (Women, Multi-ancestry, Synthetic)." Hugging Face Datasets.
+
+and relevant epigenetic literature (e.g., Horvath 2013, Horvath & Raj 2018, promoter methylation studies of BRCA1/RASSF1A/CDH1/CDKN2A, and genome-wide methylation analyses comparing African- and European-ancestry breast tumours).