BioResearch Agent — GWAS Harmonization Skill
Capability
Harmonizes GWAS summary statistics across multiple ancestries for cross-ancestry MR:
- Cross-ancestry GWAS simulation — generates per-ancestry GWAS summary statistics with
realistic allele-frequency drift (AFR < EUR < EAS/SAS/AMR), effect-size heterogeneity, and
deliberately injected data-quality issues (allele swaps, strand confusion, palindromic SNPs).
- Allele harmonization — 5-case alignment: direct match, effect-allele swap, strand flip,
palindromic SNP removal, and unmatched SNP exclusion. Standardizes all ancestries to a common
effect-allele / non-effect-allele convention.
- Strand ambiguity resolution — removes palindromic SNPs (A/T, G/C) near EAF = 0.5 where
strand orientation cannot be inferred, using a MAF-distance threshold (default ±0.01).
- Allele-frequency divergence — computes pairwise Fst-like statistics across ancestries to
quantify genetic divergence and flag SNPs with extreme AF differences (a source of MR bias).
- Cross-ancestry signal overlap — identifies genome-wide significant SNPs shared across
ancestries vs ancestry-specific, informing instrument selection and portability assessment.
Returns harmonized per-ancestry GWAS + AF comparison + overlap report, not a causal claim.
Run
bioresearch run gwas-harmonization --n-snps 500 --n-causal 30 --seed 42 --output-dir outputs/gwas-harmonization
Outputs (in --output-dir)
gh_harmonized_gwas.csv — per-ancestry harmonized GWAS (SNP, CHR, POS, EA, NEA, BETA, SE, P, EAF, N)gh_af_comparison.csv — pairwise allele-frequency differences (SNP, ancestry_pair, af_diff, fst_like)gh_cross_ancestry_overlaps.csv — genome-wide significant SNPs per ancestry + overlap countgh_harmonization_report.txt — human-readable summary (n_harmonized, n_removed, n_palindromic)gh_af_divergence_heatmap.png — pairwise Fst-like matrix across ancestriesgh_evidence_package.json — reproducible Evidence Package (provenance + parameters + grade)
Note
This skill dispatches to the framework's gwas-harmonization workflow / demo_gwas_harmonization.py.
It adds no analysis of its own; all computations run in the workflow modules. By default uses
simulated cross-ancestry GWAS with injected quality issues to validate the harmonization pipeline —
real-data deployment would use IEU OpenGWAS / GWAS Catalog summary statistics with GWAS-SSF schema.
Evidence grade is C (methodology validation). Part of Phase 3a (cross-ancestry MR, CPU-only).
1---2name: bioresearch-gwas-harmonization3description: Harmonize GWAS summary statistics across ancestries for cross-ancestry MR. Use when the user needs to align alleles, flip strands, resolve palindromic SNPs, compute allele-frequency divergence (Fst-like), or identify cross-ancestry genome-wide signal overlap. Supports GWAS-SSF schema; mock mode validates the harmonization pipeline with simulated cross-ancestry GWAS including strand confusion and allele flips.4---56# BioResearch Agent — GWAS Harmonization Skill78## Capability910Harmonizes GWAS summary statistics across multiple ancestries for cross-ancestry MR:11121. **Cross-ancestry GWAS simulation** — generates per-ancestry GWAS summary statistics with13 realistic allele-frequency drift (AFR < EUR < EAS/SAS/AMR), effect-size heterogeneity, and14 deliberately injected data-quality issues (allele swaps, strand confusion, palindromic SNPs).152. **Allele harmonization** — 5-case alignment: direct match, effect-allele swap, strand flip,16 palindromic SNP removal, and unmatched SNP exclusion. Standardizes all ancestries to a common17 effect-allele / non-effect-allele convention.183. **Strand ambiguity resolution** — removes palindromic SNPs (A/T, G/C) near EAF = 0.5 where19 strand orientation cannot be inferred, using a MAF-distance threshold (default ±0.01).204. **Allele-frequency divergence** — computes pairwise Fst-like statistics across ancestries to21 quantify genetic divergence and flag SNPs with extreme AF differences (a source of MR bias).225. **Cross-ancestry signal overlap** — identifies genome-wide significant SNPs shared across23 ancestries vs ancestry-specific, informing instrument selection and portability assessment.2425Returns harmonized per-ancestry GWAS + AF comparison + overlap report, not a causal claim.2627## Run2829```bash30bioresearch run gwas-harmonization --n-snps 500 --n-causal 30 --seed 42 --output-dir outputs/gwas-harmonization31```3233## Outputs (in `--output-dir`)3435- `gh_harmonized_gwas.csv` — per-ancestry harmonized GWAS (SNP, CHR, POS, EA, NEA, BETA, SE, P, EAF, N)36- `gh_af_comparison.csv` — pairwise allele-frequency differences (SNP, ancestry_pair, af_diff, fst_like)37- `gh_cross_ancestry_overlaps.csv` — genome-wide significant SNPs per ancestry + overlap count38- `gh_harmonization_report.txt` — human-readable summary (n_harmonized, n_removed, n_palindromic)39- `gh_af_divergence_heatmap.png` — pairwise Fst-like matrix across ancestries40- `gh_evidence_package.json` — reproducible Evidence Package (provenance + parameters + grade)4142## Note4344This skill dispatches to the framework's `gwas-harmonization` workflow / `demo_gwas_harmonization.py`.45It adds **no analysis of its own**; all computations run in the workflow modules. By default uses46**simulated cross-ancestry GWAS** with injected quality issues to validate the harmonization pipeline —47real-data deployment would use IEU OpenGWAS / GWAS Catalog summary statistics with GWAS-SSF schema.48Evidence grade is **C** (methodology validation). Part of Phase 3a (cross-ancestry MR, CPU-only).
Alim430