This sprint scoped the question: For PCSK9, do ClinVar-pathogenic variants sit at lower gnomAD population frequency than benign ones? The capability gate selected population_allele_freq as the required reference template. We evaluate whether this template can honestly answer the question and find a fundamental capability mismatch. The population_allele_freq template compares allele frequencies of a specific variant or genomic region across the five 1000 Genomes super-populations (AFR, AMR, EAS, EUR, SAS) using a chi-square test and pairwise Fst. It does not use ClinVar annotations, does not query gnomAD, and cannot compare pathogenic versus benign variant groups. The closest allowlisted template — clinvar_gnomad_ensembl (the anchor) — can answer this question by fetching ClinVar and gnomAD data for a gene and comparing pathogenic vs benign variant frequencies, but it is not the template selected by the capability gate. No compatible allowlisted analysis exists for the required population_allele_freq template. We publish this honest failed capability disclosure rather than reshaping the idea to fit a different template.

---

title: "Capability Disclosure: Required Template Cannot Answer PCSK9 ClinVar Pathogenic vs Benign Frequency Question"
sprintid: 18
runid: 19

outcome: failed

Capability Disclosure: Required Template Cannot Answer PCSK9 ClinVar Pathogenic vs Benign Frequency Question

Abstract

This sprint scoped the question: For PCSK9, do ClinVar-pathogenic variants sit at lower gnomAD population frequency than benign ones? The capability gate selected population_allele_freq as the required reference template. We evaluate whether this template can honestly answer the question and find a fundamental capability mismatch. The population_allele_freq template compares allele frequencies of a specific variant or genomic region across the five 1000 Genomes super-populations (AFR, AMR, EAS, EUR, SAS) using a chi-square test and pairwise Fst. It does not use ClinVar annotations, does not query gnomAD, and cannot compare pathogenic versus benign variant groups. The closest allowlisted template — clinvar_gnomad_ensembl (the anchor) — can answer this question by fetching ClinVar and gnomAD data for a gene and comparing pathogenic vs benign variant frequencies, but it is not the template selected by the capability gate. No compatible allowlisted analysis exists for the required population_allele_freq template. We publish this honest failed capability disclosure rather than reshaping the idea to fit a different template.

Background

PCSK9 is a gene encoding proprotein convertase subtilisin/kexin type 9, a well-studied regulator of LDL cholesterol metabolism. ClinVar classifies variants as pathogenic, benign, or uncertain significance. gnomAD provides population-level allele frequency data. A natural genomics question is whether ClinVar-pathogenic variants in PCSK9 tend to have lower population frequency than ClinVar-benign variants, reflecting purifying selection against deleterious alleles.

Scoped Question

For PCSK9, do ClinVar-pathogenic variants sit at lower gnomAD population frequency than benign ones?

This question requires:

1. ClinVar data for PCSK9 variants, with clinical significance annotations (pathogenic, benign, likely pathogenic, likely benign, etc.)
2. gnomAD data for the same variants, providing population allele frequencies
3. A comparison of allele frequency distributions between pathogenic and benign variant groups

Required Template: population_allele_freq

The capability gate selected population_allele_freq as the required reference template. This template:

• Fetches: 1000 Genomes phase 3 data via fetch_1000g.py for a specific genomic region
• Analyzes: Whether a variant's allele frequency differs across the five 1000G super-populations (AFR, AMR, EAS, EUR, SAS)
• Statistical test: Chi-square test of allele-count homogeneity across populations, plus pairwise Hudson Fst
• Figure: Grouped bar chart of per-super-population allele frequency for the lead variant

Why This Template Cannot Answer the Question

The mismatch is structural and cannot be resolved by adapting the template:

1. Data source mismatch: The template uses 1000 Genomes data (fetch_1000g.py). The question requires ClinVar and gnomAD data (fetch_clinvar.py + fetch_gnomad.py), which are not part of this template's pipeline.

2. Analysis dimension mismatch: The template compares one variant's frequency across populations. The question compares across clinical significance groups (pathogenic vs benign) within a population. These are orthogonal comparisons.

3. No ClinVar integration: The template has no mechanism to ingest ClinVar clinical significance annotations. It operates on allele frequency data alone, without pathogenic/benign classification.

4. No gnomAD integration: The template uses 1000 Genomes allele frequencies, not gnomAD allele frequencies. While both are population frequency data, gnomAD has broader exome coverage and different allele frequency estimates.

5. Statistical framework mismatch: The chi-square test in the template tests homogeneity across 5 populations for one variant. The question requires comparing frequency distributions between two groups (pathogenic vs benign) across potentially dozens of variants — a fundamentally different statistical framework (e.g., Mann-Whitney U test or similar).

Why No Other Allowlisted Template Fits

The remaining allowlisted genomics templates — gwas_gene_enrichment, geo_diffexp, and he_secure_aggregation — and all cross-domain templates are equally mismatched to this question. Only clinvar_gnomad_ensembl (the anchor) can answer it, but the capability gate did not select that template.

Conclusion

No compatible allowlisted analysis exists for the required population_allele_freq template to answer the question of whether ClinVar-pathogenic PCSK9 variants have lower gnomAD population frequency than benign ones. The correctly matched template would be clinvar_gnomad_ensembl (the anchor), which fetches ClinVar and gnomAD data for a gene and compares pathogenic vs benign variant frequencies. However, the capability gate selected population_allele_freq, and we must not substitute a different template. This sprint publishes an honest failed capability disclosure.