Upload README.md with huggingface_hub

README.md

@@ -0,0 +1,259 @@
+---
+language:
+- en
+tags:
+- tumour-markers
+- ctDNA
+- treatment-response
+- trajectories
+- breast-cancer
+- sub-saharan-africa
+license: cc-by-nc-4.0
+pretty_name: SSA Breast Biomarker Trajectories (CA15-3 / CA27-29 / ctDNA)
+task_categories:
+- other
+size_categories:
+- 1K<n<10K
+---
+
+# SSA Breast Biomarker Trajectories (Synthetic)
+
+## Dataset summary
+
+This module provides a **synthetic biomarker trajectory dataset** for breast cancer treatment monitoring, focused on:
+
+- **Serum tumour markers**: CA 15-3 and CA 27-29.
+- **Circulating tumour DNA (ctDNA)** levels.
+- **Treatment response monitoring** over 0–24 months.
+
+It is designed to emulate:
+
+- Higher baseline marker levels in advanced-stage disease.
+- Early declines in CA 15-3 and ctDNA in responders.
+- Rising or non-clearing ctDNA in early and late progression.
+
+All records are **fully synthetic**, grounded in clinical literature on serial CA15-3/CA27-29 measurement and ctDNA dynamics in metastatic and early breast cancer.
+
+## Cohort design
+
+### Sample size and populations
+
+- **Baseline patients**: 6,000.
+- **Populations**:
+  - `SSA_West`: 1,500
+  - `SSA_East`: 1,500
+  - `SSA_Central`: 1,000
+  - `SSA_Southern`: 1,000
+  - `AAW` (African American women): 1,000
+  - `EUR`: 600
+  - `EAS`: 400
+
+### Baseline variables
+
+- `sex`: predominantly `Female` (~99%).
+- `age_years`: 18–90, with slightly younger distribution in SSA vs EUR/EAS.
+- `stage_at_diagnosis`: I–IV, with later stages more frequent in SSA.
+- `tumor_subtype`:
+  - `Luminal_A`, `Luminal_B`, `HER2_enriched`, `Basal_like`, `Normal_like`, with higher basal-like in SSA and AAW.
+- `therapy_class`:
+  - `Endocrine_only`
+  - `Chemo_only`
+  - `Chemo_plus_endocrine`
+  - `HER2_targeted`
+  - `CDK4_6_plus_endocrine`
+- `response_class` (treatment course):
+  - `Deep_response`
+  - `Partial_response`
+  - `Stable_disease`
+  - `Early_progression`
+  - `Late_progression`
+
+These response classes drive **biomarker trajectories**.
+
+## Time structure
+
+Patients are followed at:
+
+- `time_months`: [0, 3, 6, 9, 12, 18, 24].
+
+At each time point, biomarker levels are simulated conditional on:
+
+- `stage_at_diagnosis` (baseline tumour burden).
+- `response_class` (kinetic pattern).
+
+## Biomarkers
+
+### CA 15-3
+
+- Units: `U/mL`.
+- Baseline means increase with stage (approximate):
+  - Stage I: ~15 U/mL
+  - Stage II: ~30 U/mL
+  - Stage III: ~60 U/mL
+  - Stage IV: ~120 U/mL
+- SD fraction ~0.6 (log-normal variability).
+
+Response-class–specific trajectories (multipliers on baseline):
+
+- `Deep_response`: rapid drop (e.g., ~50% at 3 months, ~75% reduction by 12 months).
+- `Partial_response`: moderate drop (~30% at 3 months, ~50% reduction by 12 months).
+- `Stable_disease`: flat or slightly drifting around baseline.
+- `Early_progression`: early rise (>10% increase by 3 months, continuing upward).
+- `Late_progression`: initial drop, followed by rise after ~12 months.
+
+These patterns reflect published CA15-3 kinetics linking rapid declines to radiologic response and surges to progression.
+
+### CA 27-29
+
+- Units: `U/mL`.
+- Modelled as a **correlated companion marker** to CA 15-3:
+  - `ca2729_u_per_ml` is sampled as CA 27-29 ≈ CA 15-3 × log-normal noise.
+- Captures that CA 27-29 is clinically used in a similar role to CA 15-3 but with independent assay noise.
+
+### ctDNA
+
+- Units: `copies_per_mL`.
+- Baseline log10 means increase with stage:
+  - Stage I: ~10^1.3
+  - Stage II: ~10^1.7
+  - Stage III: ~10^2.1
+  - Stage IV: ~10^2.5
+- Detection limit: ~5 copies/mL.
+
+Kinetic patterns by `response_class`:
+
+- `Deep_response`:
+  - Rapid and deep ctDNA decline; many become undetectable by 3–6 months.
+- `Partial_response`:
+  - Substantial decline but less complete clearance.
+- `Stable_disease`:
+  - Small changes, often hovering near baseline.
+- `Early_progression`:
+  - Transient modest fall, followed by strong rise (often >2–3× by 12–24 months).
+- `Late_progression`:
+  - Early decline then late rise (mirroring late progression in CA 15-3).
+
+Summary flags:
+
+- `ctdna_baseline_detectable`
+- `ctdna_cleared_by_6m` (detectable at baseline and undetectable at all timepoints ≤6 months).
+
+## Files and schema
+
+### Baseline table
+
+Files:
+
+- `biomarker_trajectories_baseline.parquet`
+- `biomarker_trajectories_baseline.csv`
+
+Columns (per patient):
+
+- Identifiers & demographics:
+  - `sample_id`, `population`, `region`, `is_SSA`, `is_reference_panel`, `sex`, `age_years`.
+- Tumour & therapy:
+  - `stage_at_diagnosis`, `tumor_subtype`, `therapy_class`, `response_class`.
+- CA 15-3 summary:
+  - `ca153_baseline_u_per_ml`
+  - `ca153_nadir_u_per_ml`
+  - `ca153_best_delta_fraction` (nadir vs baseline, negative values indicate decline).
+- CA 27-29 summary:
+  - `ca2729_baseline_u_per_ml`
+  - `ca2729_nadir_u_per_ml`
+  - `ca2729_best_delta_fraction`.
+- ctDNA summary:
+  - `ctdna_baseline_copies_per_ml`
+  - `ctdna_min_copies_per_ml`
+  - `ctdna_baseline_detectable`
+  - `ctdna_cleared_by_6m`.
+
+### Timeseries table
+
+Files:
+
+- `biomarker_trajectories_timeseries.parquet`
+- `biomarker_trajectories_timeseries.csv`
+
+Columns (per patient × timepoint):
+
+- `sample_id`
+- `population`
+- `stage_at_diagnosis`
+- `tumor_subtype`
+- `therapy_class`
+- `response_class`
+- `time_months`
+- `ca153_u_per_ml`
+- `ca2729_u_per_ml`
+- `ctdna_copies_per_ml`
+- `ctdna_detectable`
+
+## Generation
+
+The dataset is generated using:
+
+- `biomarker_trajectories/scripts/generate_biomarker_trajectories.py`
+
+with configuration:
+
+- `biomarker_trajectories/configs/biomarker_trajectories_config.yaml`
+
+and literature inventory:
+
+- `biomarker_trajectories/docs/LITERATURE_INVENTORY.csv`
+
+Key steps:
+
+1. **Baseline cohort**: draw populations, sex, age, stage, subtype, therapy, and response class.
+2. **Baseline markers**: sample CA 15-3 and ctDNA baseline values by stage using log-normal distributions.
+3. **Trajectories**: apply response-class–specific multipliers at each timepoint, with log-normal noise, to generate CA 15-3, CA 27-29, and ctDNA levels.
+4. **Summaries**: compute nadirs, best percentage changes, baseline detectability, and ctDNA clearance.
+
+## Validation
+
+Validation is performed with:
+
+- `biomarker_trajectories/scripts/validate_biomarker_trajectories.py`
+
+and summarised in:
+
+- `biomarker_trajectories/output/validation_report.md`
+
+Checks include:
+
+- **C01–C02**: Sample size and population counts vs configuration.
+- **C03**: CA 15-3 baseline means increase monotonically with stage.
+- **C04**: CA 15-3 kinetics by response class (deep responders show early declines; early progressors show rises).
+- **C05**: ctDNA clearance is higher in deep responders and low in progression classes.
+- **C06**: Baseline ctDNA detectability is substantially higher in stage IV than in stage I.
+- **C07**: Missingness in key baseline and timeseries variables.
+
+## Intended use
+
+This dataset is intended for:
+
+- Developing and benchmarking models for **biomarker-based response assessment**.
+- Exploring **surrogate endpoints** based on CA 15-3, CA 27-29, and ctDNA dynamics.
+- Teaching about biomarker kinetics in metastatic and advanced breast cancer.
+
+It is **not intended** for:
+
+- Estimating real-world cutoffs or predictive values.
+- Guiding individual patient management.
+
+## Ethical considerations
+
+- All data are **synthetically generated** from literature-informed parameters.
+- Population and clinical patterns are inspired by published cohorts but do not represent any specific registry or trial.
+- The goal is to support method development and education around biomarker trajectories in diverse breast cancer populations, including SSA.
+
+## License
+
+- License: **CC BY-NC 4.0**.
+- Free for non-commercial research, method development, and education with attribution.
+
+## Citation
+
+If you use this dataset, please cite:
+
+> Electric Sheep Africa. "SSA Breast Biomarker Trajectories (CA15-3 / CA27-29 / ctDNA, Synthetic)." Hugging Face Datasets.