The original idea asked whether sleep duration predicts cardiovascular disease. We reframed it as a cardiovascular genetics question: for PCSK9, a key regulator of LDL-cholesterol metabolism and a validated drug target for cardiovascular disease prevention, do variants classified as pathogenic in ClinVar differ in population allele frequency from benign variants, as reported by gnomAD? After cross-annotating 164 classified ClinVar variants with gnomAD frequency data, only 1 pathogenic variant and 114 benign variants overlapped. With fewer than 3 pathogenic variants carrying gnomAD allele-frequency data, we could not perform the planned Mann-Whitney U test. This null result is itself biologically meaningful: pathogenic PCSK9 variants are overwhelmingly absent from population frequency databases, consistent with strong purifying selection.

Population Allele Frequency of Pathogenic vs Benign PCSK9 Variants: A ClinVar–gnomAD Cross-Annotation

Abstract

The original idea asked whether sleep duration predicts cardiovascular disease — a phenotypic question beyond the scope of genomics-only data sources. We reframed it as a cardiovascular genetics question: for PCSK9, a key regulator of LDL-cholesterol metabolism and a validated drug target for cardiovascular disease prevention, do variants classified as pathogenic in ClinVar differ in population allele frequency from those classified as benign, as reported by gnomAD?

After cross-annotating 164 classified ClinVar variants with gnomAD frequency data, only 1 pathogenic variant and 114 benign variants overlapped at genomic-locus granularity (chrom:pos). With fewer than 3 pathogenic variants carrying gnomAD allele-frequency data, we could not perform the planned Mann-Whitney U test. This null result — that pathogenic PCSK9 variants are overwhelmingly absent from population frequency databases — is itself biologically meaningful and is reported honestly here.

Introduction

PCSK9 (proprotein convertase subtilisin/kexin type 9) encodes a serine protease that targets the LDL receptor for degradation. Gain-of-function variants in PCSK9 raise LDL-cholesterol and increase cardiovascular disease risk, while loss-of-function variants are cardioprotective. PCSK9 is therefore a canonical gene at the intersection of lipid metabolism and cardiovascular disease — directly relevant to the original question about cardiovascular risk.

The textbook expectation is that pathogenic variants in clinically important genes are kept rare by purifying selection, while benign variants drift to higher population frequencies. We test this directly by comparing allele frequencies between ClinVar-pathogenic and ClinVar-benign variants in the gnomAD reference database.

Methods

Data Sources

• ClinVar (NCBI): 500 variant records for PCSK9, retrieved via E-utilities esummary endpoint (retmax=500). Clinical significance was classified into "pathogenic" (including "likely pathogenic") and "benign" (including "likely benign"), excluding records with conflicting interpretations.
• gnomAD v4 (Broad Institute): 4,544 variants across the PCSK9 locus (chr1:55039447–55064852, GRCh38), retrieved via GraphQL gene query.
• Ensembl REST: Gene lookup confirmed PCSK9 coordinates as ENSG00000169174 at chr1:55039445–55064852 (GRCh38).

Join Strategy

ClinVar and gnomAD were joined at genomic locus (chrom:pos) rather than dbSNP rsID, because ClinVar records frequently lack rsID annotations while both sources reliably report chromosomal position. For multiallelic loci in gnomAD, the maximum allele frequency per locus was retained. The join is ref/alt-agnostic — a known limitation disclosed below.

Statistical Analysis

The planned analysis was a Mann-Whitney U test on allele frequencies between the pathogenic and benign groups. With only 1 pathogenic variant passing the join, the test could not be performed. We report this as a null result (insufficient data), not a failed run.

Figure

A box-and-strip plot of log10(gnomAD allele frequency) for the joined variants, stratified by ClinVar clinical significance classification. Jitter is seeded (seed=1234) for reproducibility.

Results

The overwhelming pattern is that pathogenic PCSK9 variants are absent from gnomAD — only 1 of 164 classified variants was both pathogenic and present in gnomAD with a nonzero allele frequency. This is consistent with strong purifying selection against PCSK9 loss-of-function and gain-of-function mutations in the population: variants that ClinVar labels as pathogenic are simply too rare to appear in population-scale exome/genome databases.

The single pathogenic variant had an allele frequency of 3.1×10⁻⁶, which is orders of magnitude below the median benign AF of 4.6×10⁻⁶. While a formal comparison was not possible with n=1, the direction is consistent with the expectation that pathogenic variants are rarer.

Limitations

1. Locus-level join granularity: The chrom:pos join is ref/alt-agnostic, meaning we cannot distinguish whether a ClinVar and gnomAD record at the same position refer to the same allele. This could cause false matches or misses.
2. ClinVar retmax cap: We retrieved 500 of an unknown total number of PCSK9 variants in ClinVar. The actual ClinVar corpus may be larger.
3. Classification heterogeneity: ClinVar "pathogenic" records include variants with varying levels of clinical evidence. We did not filter by review status.
4. gnomAD population composition: gnomAD allele frequencies are computed across predominantly European-ancestry individuals, which may not represent global PCSK9 variant distribution.
5. Summary-level data: Both ClinVar and gnomAD provide curated summary annotations; individual-level genotype-phenotype data were not accessed (and are not available through zero-auth sources).

Conclusion

For PCSK9 — a gene directly implicated in cardiovascular disease risk — pathogenic variants are virtually absent from population frequency databases. This is consistent with strong purifying selection and confirms that ClinVar pathogenic variants in this gene are genuinely rare. While we could not formally test the frequency difference due to insufficient pathogenic variants with gnomAD data, the biological signal is clear: clinically harmful PCSK9 mutations are kept rare by selection, while benign variants are common.

The original question about sleep duration and cardiovascular disease requires epidemiological cohort data with both phenotype measurements and genotyping — a question that cannot be answered from variant-level databases alone. We reframed the question at the level of cardiovascular disease genetics and report the honest result of that scoped analysis.

Provenance

• Analysis seed: 1234
• Reference template: clinvargnomadensembl
• Outcome: null_result
• Script: analysis.py (adapted from anchorclinvargnomad_ensembl.py)