electricsheepafrica/africa-synth-drug-metabolism-pharmacogenomics-congo-rep

⚠️ Synthetic dataset — Parameterized from published SSA literature, not real observations. Not suitable for empirical analysis or policy inference.

SSA Drug Metabolism Pharmacogenomics (Oncology-focused, Synthetic)

Dataset summary

This dataset provides a synthetic pharmacogenomics panel for drug metabolism genes relevant to oncology, with a strong focus on sub-Saharan African (SSA) populations and African diaspora. It models genotype-derived metabolizer phenotypes and dose-group recommendations for key genes:

• CYP2D6 – tamoxifen and multiple other drugs.
• CYP3A5 – metabolism of several chemotherapies and immunosuppressants.
• TPMT – thiopurine drugs.
• UGT1A1 – irinotecan toxicity.
• DPYD – 5-fluorouracil / capecitabine toxicity.

All individuals and genotypes are fully synthetic and constructed from literature-informed phenotype distributions and guideline-based dosing logic. The resource is intended for methods development, education, and workflow benchmarking in pharmacogenomics, not for clinical decision-making.

Cohort design

Sample size and populations

• Total N: 10,000 synthetic individuals.

• Populations (aligned with other Electric Sheep Africa cohorts):

  • SSA_West: 2,000
  • SSA_East: 2,000
  • SSA_Central: 1,500
  • SSA_Southern: 1,500
  • AAW (African American women, admixed): 1,500
  • EUR (European reference): 1,000
  • EAS (East Asian reference): 500

• Sex distribution:

  • Male: 50%
  • Female: 50%

These groups mirror the ancestry structure used in other SSA-focused synthetic datasets (e.g., SNP array, structural variation, colorectal/ovarian genomic), enabling cross-dataset method development.

Genes and pharmacogenomic modeling

The dataset encodes phenotypes and dose-group recommendations rather than raw variant calls, to simplify downstream use while retaining clinically meaningful structure.

CYP2D6 (tamoxifen and other substrates)

• Modeled phenotypes (CPIC-style):

  • UM – ultrarapid metabolizer.
  • NM – normal metabolizer.
  • IM – intermediate metabolizer.
  • PM – poor metabolizer.

• Phenotype distributions are approximated per population using:

  • Global CYP2D6 allele/phenotype summaries (NCBI Medical Genetics).
  • African-focused CYP2D6 polymorphism studies (high *17 frequency, low PM).

  Example (approximate) phenotype frequencies:

  • SSA (West/East/Central/Southern): UM ~3–4%, NM ~66–68%, IM ~26–28%, PM ~2%.
  • AAW: UM ~3%, NM ~67%, IM ~26%, PM ~4%.
  • EUR: UM ~2%, NM ~70%, IM ~20%, PM ~8%.
  • EAS: UM ~1%, NM ~50%, IM ~40%, PM ~9%.

• Representative diplotypes are simulated for each phenotype category (e.g. *1/*1 for NM, *4/*4 for PM, *1/*17 for IM), but are illustrative synthetic star allele labels.

• Tamoxifen recommendation (simplified CPIC logic):

  • UM: standard dose or consider lower if toxicity.
  • NM: standard dose.
  • IM: consider alternative or closer monitoring.
  • PM: consider alternative endocrine therapy.

CYP3A5 (expressor vs non-expressor)

• Phenotypes:

  • expressor – at least one functional allele.
  • non_expressor – two loss-of-function alleles.

• Guided by CYP3A5 diversity in Africa (Ingelman-Sundberg et al., BMC Genet) and other global studies:

  • African populations show ~40–55% expressors on average, but with marked intra-African variability.
  • European populations have low expressor frequency (~10%).

• The dataset encodes:

  • Higher expressor fraction in SSA and AAW.
  • Low expressor fraction in EUR; intermediate in EAS.

TPMT (thiopurine metabolism)

• Phenotypes:

  • normal – normal metabolizer.
  • intermediate – intermediate metabolizer.
  • poor – poor metabolizer.

• Based on classic TPMT activity distribution:

  • ~89–94% normal, ~5–11% intermediate, ~0.3% poor.
  • Allele distributions differ by ancestry (e.g. TPMT3C more common among African-Americans; TPMT3A dominates in Europeans).

• Dose-group mapping (simplified CPIC TPMT/NUDT15 guideline):

  • normal: standard thiopurine dose.
  • intermediate: substantial dose reduction.
  • poor: avoid or drastically reduce thiopurines.

UGT1A1 (irinotecan toxicity)

• Modeled as risk categories combining UGT1A1*28 and *6 effects:

  • normal – low irinotecan toxicity risk.
  • intermediate – modestly increased risk.
  • high_risk – substantially increased risk.

• Informed by:

  • Irinotecan/UGT1A1 Medical Genetics summaries.
  • Asian PGx review (Lo et al., 2020): UGT1A1*6 MAF ~14.6% in Asians vs ~0.8% in Caucasians.

• Risk distributions by population approximate:

  • SSA & AAW: higher fraction of intermediate and high_risk compared with EUR.
  • EUR: more normal, fewer high_risk than SSA.
  • EAS: elevated intermediate risk via *6, but high_risk remains relatively rare.

• Irinotecan dose group (simplified, irinotecan-analogous CPIC/DPWG logic):

  • normal: standard dose.
  • intermediate: consider reduced starting dose.
  • high_risk: substantial dose reduction or alternative regimen.

DPYD (fluoropyrimidines: 5-FU / capecitabine)

• Phenotypes (CPIC-style):

  • normal – normal DPD activity.
  • intermediate – reduced activity (heterozygous for high-risk variants).
  • poor – near-absent activity (homozygous or compound heterozygous).

• Anchored to:

  • CPIC guideline for DPYD and fluoropyrimidine dosing (2017 update).
  • European cancer-cohort variant frequencies (e.g., ~4.9% carriers of at least one high-risk DPYD variant; HapB3 ~2.9%, c.2846A>T ~1.4%, *2A ~0.7%).

• The dataset encodes higher intermediate/poor phenotypes in EUR vs SSA/EAS, with:

  • normal ~95%, intermediate ~4.5%, poor ~0.5% in EUR.
  • Lower intermediate/poor frequencies in SSA and EAS.

• Fluoropyrimidine dose group (simplified CPIC logic):

  • normal: standard dose.
  • intermediate: ~50% starting dose with careful titration.
  • poor: avoid fluoropyrimidines or use an alternative.

File and schema

1. drug_metabolism_data.parquet / drug_metabolism_data.csv

One row per synthetic individual.

Core columns:

• Demographics / ancestry

  • sample_id – synthetic ID.
  • population – one of SSA_West, SSA_East, SSA_Central, SSA_Southern, AAW, EUR, EAS.
  • region – SSA subregion or Non_SSA.
  • is_SSA – boolean.
  • is_reference_panel – True for AAW/EUR/EAS.
  • sex – Male or Female.

• CYP2D6

  • CYP2D6_phenotype – UM, NM, IM, PM.
  • CYP2D6_diplotype – illustrative star-allele diplotype (synthetic labels).

• CYP3A5

  • CYP3A5_phenotype – expressor or non_expressor.
  • CYP3A5_diplotype – illustrative diplotype (e.g. *1/*3, *3/*3).

• TPMT

  • TPMT_phenotype – normal, intermediate, poor.
  • TPMT_diplotype – synthetic TPMT star-allele diplotype.

• UGT1A1

  • UGT1A1_risk_category – normal, intermediate, high_risk.
  • UGT1A1_genotype_class – synthetic genotype class (e.g. 6/6, 6/7, 28/28).

• DPYD

  • DPYD_phenotype – normal, intermediate, poor.
  • DPYD_variant_status – high-level variant status (no_risk_variant, heterozygous_high_risk_variant, homozygous_or_compound_heterozygous_high_risk).

• Drug-specific recommendations (derived from gene phenotypes):

  • tamoxifen_recommendation – CPIC-style category based on CYP2D6_phenotype.
  • thiopurine_dose_group – dose group from TPMT_phenotype.
  • irinotecan_dose_group – dose group from UGT1A1_risk_category.
  • fluoropyrimidine_dose_group – dose group from DPYD_phenotype.

Generation

The dataset was generated with:

• drug_metabolism/scripts/generate_drug_metabolism.py

using configuration in:

• drug_metabolism/configs/drug_metabolism_config.yaml

and literature curated in:

• drug_metabolism/docs/LITERATURE_INVENTORY.csv

Key steps:

1. Sample generation
   • 10,000 individuals allocated by population and sex according to the config.
2. Pharmacogene phenotype assignment
   • For each gene, per-population phenotype/risk distributions are sampled from the config (informed by allele/phenotype frequencies from the literature inventory).
3. Synthetic diplotype/genotype class assignment
   • Representative diplotypes or genotype classes are assigned per phenotype category for interpretability.
4. Dose-group derivation
   • Drug-specific recommendations are assigned by deterministic mapping from phenotype/risk category to dose group, guided by CPIC-like logic.

Validation

Validation follows the GENOMICS Synthetic Data Playbook and is implemented in:

• drug_metabolism/scripts/validate_drug_metabolism.py

Checks include:

• C01 – Sample size matches config
  • Confirms N = 10,000.
• C02 – Population sample sizes vs config
  • Per-population counts within ±10% relative deviation.
• C03 – Pharmacogene phenotype/risk distributions vs config
  • For each gene and population, compares observed phenotype/risk frequencies to configured targets.
• C04 – Dose-group consistency
  • Verifies that dose-group fields (tamoxifen_recommendation, thiopurine_dose_group, irinotecan_dose_group, fluoropyrimidine_dose_group) are consistent with their driving phenotypes/risk categories.
• C05 – Missingness in key variables
  • Ensures negligible missingness in demographics, gene phenotypes, genotype classes, and dose-group recommendations.

The validator outputs a Markdown report:

• drug_metabolism/output/validation_report.md

For the released version, all checks complete with an overall status of PASS.

Intended use

This dataset is intended for:

• Methods development and benchmarking in pharmacogenomics (PGx), focusing on:
  • Phenotype calling from genotype.
  • Dose-group assignment logic.
  • Ancestry-aware PGx modeling.
• Teaching and demonstration of PGx concepts in oncology, including:
  • Metabolizer phenotypes.
  • Gene–drug interactions.
  • Dose adjustments for toxicity risk.

The dataset is not suitable for:

• Clinical decision-making or patient care.
• Individual-level risk prediction.
• Direct inference about real-world groups.

All data are synthetic and do not represent any actual person.

Ethical considerations

• No real patient data are used.
• Population labels (SSA subregions, AAW, EUR, EAS) are for simulation realism only.
• Analyses and visualizations should be framed as methodological or educational, not as definitive statements about real populations.

License

• License: CC BY-NC 4.0.
• Free to use for non-commercial research, teaching, and methods development with proper attribution.

Citation

If you use this dataset, please cite:

Electric Sheep Africa. "SSA Drug Metabolism Pharmacogenomics (Oncology-focused, Synthetic)." Hugging Face Datasets.

and consider citing the core PGx resources that informed the design, such as:

• Johansson I et al. CYP2D6 allele frequency in European Caucasians, Asians, Africans and their descendants.
• Ingelman-Sundberg M et al. Molecular diversity and population structure at the CYP3A5 gene in Africa.
• Weinshilboum R et al. Polymorphism of the Thiopurine S-Methyltransferase Gene in Caucasians and African-Americans.
• Lo C et al. Pharmacogenomics in Asian subpopulations and impacts on commonly prescribed medications.
• Caudle KE et al. CPIC Guideline for Thiopurines and TPMT and NUDT15.
• Amstutz U et al. CPIC Guideline for Dihydropyrimidine Dehydrogenase Genotype and Fluoropyrimidine Dosing.