COMPUTE, DON'T DESCRIBE

• Allele frequencies from gnomAD -> Use tooluniverse-population-genetics
• ClinVar / OMIM variant interpretation -> Use tooluniverse-variant-interpretation
• GWAS fine-mapping -> Use tooluniverse-gwas-finemapping

Phase 1: Search 1000 Genomes Populations

IGSR_search_populations: superpopulation (string/null, one of AFR/AMR/EAS/EUR/SAS), query (string/null, free-text search by name), limit (int). Returns {status, data: {total, populations: [{code, name, description, sample_count, superpopulation_code, superpopulation_name, latitude, longitude}]}, metadata: {source, filter_superpopulation, filter_query}}.

Superpopulation codes:

// List all AFR populations
{"superpopulation": "AFR", "limit": 10}

// Search by name (free-text)
{"query": "Yoruba", "limit": 5}

// List all populations
{"limit": 26}

Response example:

{
  "status": "success",
  "data": {
    "total": 3,
    "populations": [
      {"code": "YRI", "name": "Yoruba", "description": "Yoruba in Ibadan, Nigeria",
       "sample_count": 188, "superpopulation_code": "AFR", "superpopulation_name": "African Ancestry"}
    ]
  }
}

Phase 2: Search Samples by Population

IGSR_search_samples: population (string/null, population code e.g. "YRI"), data_collection (string/null, collection title), sample_name (string/null, specific sample e.g. "NA12878"), limit (int). Returns {status, data: {total, samples: [{name, sex, biosample_id, populations: [{code, name, superpopulation}], data_collections: [...]}]}}.

// Find all YRI samples
{"population": "YRI", "limit": 10}

// Look up the reference sample NA12878
{"sample_name": "NA12878", "limit": 1}

// Find samples in the 30x high-coverage collection
{"data_collection": "1000 Genomes 30x on GRCh38", "limit": 5}

NOTE: population takes a population code (e.g. "YRI", "GBR", "CHB"), not a superpopulation code. Use IGSR_search_populations first to get population codes if starting from a superpopulation.

Phase 3: List Data Collections

IGSR_list_data_collections: limit (int). Returns {status, data: {total, collections: [{code, title, short_title, sample_count, population_count, data_types, website}]}}.

{"limit": 20}

Key collections available (18 total):

Phase 4: GWAS Context for Population Stratification

Search GWAS associations for a trait

gwas_search_associations: trait (string, free text), limit (int). Returns GWAS associations with rsID, p-value, mapped genes, EFO trait IDs.

{"trait": "type 2 diabetes", "limit": 10}

Get variants for a specific trait (by EFO ID)

gwas_get_variants_for_trait: trait (string, EFO ID e.g. "EFO_0001645"), limit (int).

{"trait": "EFO_0001645", "limit": 10}

Find SNPs in a gene from GWAS catalog

gwas_get_snps_for_gene: gene_symbol (string), limit (int). Returns SNPs mapped to the gene with rsIDs, genomic positions, functional classes.

{"gene_symbol": "TCF7L2", "limit": 10}

Workflow: Population Stratification in GWAS

Step 1 -- Find populations of interest:

// Get all EUR populations
{"superpopulation": "EUR", "limit": 10}
// -> Returns codes like GBR, FIN, CEU, TSI, IBS

Step 2 -- Get samples from target population:

// Get YRI samples (AFR)
{"population": "YRI", "limit": 100}

Step 3 -- Get GWAS SNPs for the gene or trait:

// GWAS hits for TCF7L2 (T2D gene)
{"gene_symbol": "TCF7L2", "limit": 20}

Step 4 -- Cross-reference with population data for stratification analysis.

Common Population Codes

Reasoning Framework for Result Interpretation

Evidence Grading

Interpretation Guidance

• Allele frequency interpretation by ancestry: Allele frequencies vary across superpopulations (AFR, AMR, EAS, EUR, SAS) due to genetic drift, selection, and demographic history. AFR populations have the highest genetic diversity and longest haplotypes broken by recombination. Disease-risk alleles may be common in one ancestry and rare in another, leading to differential genetic risk across populations.
• Fst significance thresholds: Fst measures population differentiation (0 = no differentiation, 1 = complete fixation of different alleles). Global Fst for human populations averages ~0.12. Locus-specific Fst > 0.3 suggests strong differentiation (possible selection). Fst > 0.5 is extreme and rare in humans outside known selection targets (e.g., SLC24A5 for skin pigmentation). Compare locus Fst against genome-wide distribution to identify outliers.
• LD interpretation: Linkage disequilibrium (LD) patterns differ by ancestry. AFR populations have shorter LD blocks due to older demographic history, requiring denser genotyping for fine-mapping. EUR and EAS populations have longer LD blocks. When a GWAS hit is in LD with multiple variants, the causal variant is more likely to be resolved in AFR-ancestry data. Report r-squared values: r2 > 0.8 = strong LD, 0.2-0.8 = moderate, < 0.2 = weak.
• Population stratification: Uncontrolled population structure in GWAS inflates false positives. The 1000 Genomes superpopulation labels provide a framework for stratified analysis. Mixed-ancestry samples (e.g., AMR) require local ancestry deconvolution for accurate interpretation.
• Sample size context: 1000 Genomes has ~2500 samples across 26 populations. Population-specific allele frequencies have limited precision for smaller populations (N < 100). For rare variants (AF < 0.01), larger resources like gnomAD provide more reliable estimates.

Synthesis Questions

1. Does the allele frequency of the variant of interest differ meaningfully (> 5%) across superpopulations, and could this explain differential disease prevalence or GWAS effect sizes?
2. Is the GWAS association replicated across ancestries, or is it population-specific, potentially due to LD structure differences or population-specific selection?
3. For fine-mapping, does the LD pattern in AFR populations narrow the association signal compared to EUR, helping identify the likely causal variant?
4. Are the population labels and sample sizes in the 1000 Genomes dataset adequate for the analysis, or is the target population underrepresented?
5. Could population stratification (uncontrolled ancestry differences between cases and controls) explain the observed association, rather than a true genetic effect?

Tool Parameter Quick Reference