electricsheepafrica/ssa-breast-gut-microbiome

SSA Breast Gut Microbiome Dataset (Women, Multi-ancestry, Synthetic)

Dataset summary

This dataset provides a synthetic gut microbiome cohort linked conceptually to women with invasive breast cancer across multiple ancestry groups, with emphasis on Sub-Saharan Africa (SSA) and comparable reference populations.

It approximates 16S rRNA stool profiles at a coarse genus/functional level, focusing on factors that vary with diet and geography:

• Diet patterns – high-fiber plant-rich vs mixed-transition vs Westernized diets.
• Enterotypes – Bacteroides-, Prevotella-, and Ruminococcus-like communities.
• Alpha diversity – Shannon index and observed richness.
• Core genus relative abundances – Bacteroides, Prevotella, Faecalibacterium, Ruminococcus.
• Functional scores – predicted potential for short-chain fatty acid (SCFA) production, bile acid transformation, and mucin degradation, loosely inspired by PICRUSt/KEGG/MetaCyc pathway analyses.

The dataset is informed by De Filippo et al. (Burkina Faso vs Italy), Yatsunenko et al. (age and geography), Arumugam et al. (enterotypes), SSA rural/urban cohorts, and functional prediction studies, while remaining fully synthetic and non-identifiable.

Cohort design

Sample size and populations

• Total N: 10,000 synthetic gut microbiome 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

Demographic structure mirrors other Electric Sheep Africa breast cancer modules:

• Predominantly women, small male fraction (~1%).
• Ages 18–90, with younger mean ages in SSA vs older in EUR/EAS/AAW.

Diet patterns and enterotypes

Diet pattern

Variable:

• diet_pattern – one of:
  • High_fiber_plant_rich
  • Mixed_transition
  • Westernized_high_fat_sugar

Anchoring:

• High_fiber_plant_rich – inspired by rural African and Amerindian diets (high fiber, resistant starch, low animal fat) in De Filippo and Yatsunenko.
• Westernized_high_fat_sugar – reflects urban/industrialized diets rich in animal protein, fat, and refined carbohydrates.
• Mixed_transition – intermediate pattern seen in migrants and urbanizing SSA cohorts.

SSA populations are enriched for High_fiber_plant_rich and Mixed_transition patterns, while EUR/EAS have higher probabilities of Westernized_high_fat_sugar diets.

Enterotypes

Variable:

• enterotype – Bacteroides, Prevotella, or Ruminococcus_like.

Anchoring:

• Based on Arumugam et al. and later enterotype literature.
• Prevotella enterotypes are more common in high-fiber, plant-rich diets (rural SSA and some AAW groups).
• Bacteroides enterotypes are enriched in Westernized diets (EUR/EAS, some AAW urban cohorts).
• Ruminococcus-like enterotypes represent intermediate/alternative communities.

In this dataset:

• SSA_West/East/Central have Prevotella fractions ≈0.50–0.55 and Bacteroides ≈0.25–0.30.
• EUR/EAS have Bacteroides ≈0.52–0.60 and Prevotella ≈0.15–0.20.

Diversity metrics

Alpha diversity

Variables:

• shannon_index – Shannon diversity index (log base e).
• observed_richness – approximate number of observed taxa.

Anchoring:

• Yatsunenko et al. and rural/urban SSA studies report higher richness and Shannon diversity in traditional, high-fiber populations vs Western cohorts.

Approximate configuration:

• SSA_West/East/Central: Shannon ≈4.0–4.1; richness ≈185–190.
• SSA_Southern/AAW: slightly lower but still elevated.
• EUR/EAS: Shannon ≈3.4–3.5; richness ≈155–160.

Beta diversity

This dataset does not pre-compute beta diversity matrices (e.g., UniFrac, Bray–Curtis). Instead, it provides per-sample genus-level profiles (rel_abundance_* columns) and diversity metrics that allow users to compute:

• Ordination (PCoA) on genus relative abundances.
• Distance matrices (Bray–Curtis, Jaccard, etc.).

Core genus profiles

Variables (relative abundances):

• rel_abundance_Bacteroides
• rel_abundance_Prevotella
• rel_abundance_Faecalibacterium
• rel_abundance_Ruminococcus

Construction:

• For each enterotype, a template profile is defined (e.g., Bacteroides-dominant vs Prevotella-dominant), with small per-sample Gaussian noise followed by renormalization.
• This mirrors 16S-derived community structures without modeling hundreds of taxa.

Examples:

• Bacteroides enterotype: high Bacteroides, low Prevotella.
• Prevotella enterotype: high Prevotella, lower Bacteroides.
• Ruminococcus-like: more balanced Bacteroides/Prevotella with increased Ruminococcus and Faecalibacterium.

Functional pathway scores

Variables:

• functional_SCFA_potential – predicted short-chain fatty acid (e.g., butyrate) production capacity.
• functional_Bile_acid_potential – predicted bile acid transformation capacity.
• functional_Mucin_deg_potential – predicted mucin-degradation capacity.

Interpretation:

• These are dimensionless scores on a 0–5 scale, intended to mimic PICRUSt/KEGG/MetaCyc-inferred functional potential from 16S data.
• Means are enterotype-specific:
  • Prevotella enterotypes have higher SCFA potential and somewhat lower bile acid potential.
  • Bacteroides enterotypes show relatively higher bile acid potential.
  • Ruminococcus-like enterotypes show intermediate SCFA and bile acid, with slightly higher mucin-degradation potential.

File and schema

gut_microbiome_data.parquet / gut_microbiome_data.csv

Each row represents a synthetic gut microbiome profile:

• Demographics

  • sample_id
  • population
  • region
  • is_SSA
  • is_reference_panel
  • sex
  • age

• Diet and enterotype

  • diet_pattern
  • enterotype

• Alpha diversity

  • shannon_index
  • observed_richness

• Core genera

  • rel_abundance_Bacteroides
  • rel_abundance_Prevotella
  • rel_abundance_Faecalibacterium
  • rel_abundance_Ruminococcus

• Functional scores

  • functional_SCFA_potential
  • functional_Bile_acid_potential
  • functional_Mucin_deg_potential

Generation

The dataset is generated using:

• gut_microbiome/scripts/generate_gut_microbiome.py

with configuration in:

• gut_microbiome/configs/gut_microbiome_config.yaml

and literature inventory in:

• gut_microbiome/docs/LITERATURE_INVENTORY.csv

Generation steps:

1. Cohort sampling – multi-ancestry demographic cohort.
2. Diet pattern assignment – population-specific probabilities reflecting rural/traditional, transitional, and Westernized diets.
3. Enterotype assignment – population-specific Bacteroides/Prevotella/Ruminococcus-like distributions conditioned on geography and diet literature.
4. Alpha diversity sampling – Shannon and richness drawn by population, matching rural vs Western richness trends.
5. Genus profile sampling – sample core genus relative abundances from enterotype-specific templates with noise and renormalization.
6. Functional scores – draw SCFA, bile acid, and mucin-degradation scores by enterotype.

Validation

Validation is performed with:

• gut_microbiome/scripts/validate_gut_microbiome.py

and summarized in:

• gut_microbiome/output/validation_report.md

Checks include:

• C01–C02 – Sample size and population counts vs config.
• C03 – Diet pattern distributions by population.
• C04 – Enterotype distributions by population.
• C05 – Alpha diversity means (Shannon and richness) by population.
• C06 – Functional score means by enterotype.
• C07 – Missingness across key variables.

All checks are configured with a 10% relative deviation tolerance, aiming for an overall status of PASS in the released version.

Intended use

This dataset is intended for:

• Epidemiologic and methodological simulations linking gut microbiome patterns with diet, geography, and breast cancer.
• Integration with other Electric Sheep Africa synthetic datasets (pathology, IHC, immune profiles, systemic markers) to build multi-modal models.
• Educational use for understanding enterotypes, diversity, and functional potential across global populations.

It is not suitable for:

• Estimating exact taxon abundances or functional pathway prevalence in any specific cohort.
• Direct clinical microbiome interpretation or individualized dietary recommendations.

Ethical considerations

• No real stool or sequencing data are used; all profiles are simulated from literature-informed distributions.
• Modeled geographic/dietary differences are intended for methodological realism and should not be used to stereotype populations.

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 Gut Microbiome Dataset (Women, Multi-ancestry, Synthetic)." Hugging Face Datasets.

and, as appropriate, key literature on diet, geography, and the human gut microbiome (e.g., De Filippo et al. 2010, Yatsunenko et al. 2012, Arumugam et al. 2011, SSA rural/urban microbiome studies, and PICRUSt-based functional analyses).