X-Chromosome Missense Variants Are 11.1 Percentage-Points More Likely to Be Pathogenic Than Autosomal Missense Variants in ClinVar: 47.84% Pathogenic Fraction (Wilson 95% CI [47.15, 48.54]) Across 19,965 X-Chromosome Records vs 36.72% [36.55, 36.90] Across 288,548 Autosomal Records

Abstract

We compute the per-chromosome-class Pathogenic-fraction comparison of ClinVar missense single-nucleotide variants restricted to autosomes (chromosomes 1–22), the X chromosome, the Y chromosome, and mitochondrial DNA (MT). For each variant: extract chromosomal location from the HGVS-style _id field; restrict to missense (aa.alt ≠ X); annotate via dbNSFP v4 (Liu et al. 2020) / MyVariant.info (Wu et al. 2021). Result: per-chromosome-class Pathogenic fractions:

The X-chromosome Pathogenic fraction (47.84%) exceeds the autosomal Pathogenic fraction (36.72%) by 11.12 percentage points — the Wilson 95% CIs are non-overlapping by ~10 percentage points. The X-chromosome P-fraction is 30% higher in relative terms than the autosomal baseline. The mechanism is X-linked-recessive-disorder ascertainment: X-chromosome variants are extensively curated for X-linked recessive disorders (Duchenne / Becker muscular dystrophy in DMD; hemophilia A in F8 and B in F9; X-linked intellectual disability genes; X-linked retinitis pigmentosa in RPGR / RP2; ornithine transcarbamylase deficiency in OTC; Wiskott-Aldrich syndrome in WAS; Hunter syndrome in IDS; Fabry disease in GLA; many others). X-linked recessive variants in males are hemizygous (single allele expressing the phenotype), making clinical detection sensitive; clinical-research focus on X-linked Mendelian disorders is correspondingly intense. The 47.84% X-Pathogenic fraction is also reflected in the high per-X-gene clinical-curation density: X-linked classical disease genes (DMD, F8, OTC, GLA, IDS) are among the most-densely-curated genes in ClinVar. For variant-prioritization pipelines: a missense variant on the X chromosome carries a higher Pathogenic prior than the same variant on an autosome — the per-chromosome-class prior should be applied as a metadata feature.

1. Background

X-linked Mendelian disorders are clinically distinctive:

• Recessive expression in males: males are hemizygous for X-chromosome genes; X-linked-recessive variants express phenotype with a single mutant allele (e.g., Duchenne MD, hemophilia, color blindness).
• Dominant expression in females: X-linked-dominant variants express in heterozygous females, with milder phenotypes due to X-inactivation mosaicism (e.g., Rett syndrome MECP2; X-linked hypophosphatemia).
• Heterozygous-female-carrier identification: family-history screening identifies carrier females and at-risk male offspring.

Clinical-research focus on X-linked disorders has been intensive since the early 20th century (Bell & Haldane 1937 first quantified X-linkage in hemophilia and color blindness). ClinVar X-chromosome submissions are correspondingly enriched with curated Pathogenic variants from X-linked disease cases.

This paper measures the per-chromosome-class Pathogenic-fraction asymmetry directly with formal Wilson 95% confidence intervals.

2. Method

2.1 Data

• 178,509 Pathogenic + 194,418 Benign ClinVar single-nucleotide variants from MyVariant.info, with dbNSFP v4 annotation.
• For each variant: extract chromosome from the _id HGVS-style field via regex ^chr([0-9XYMT]+). Variants on alt-contigs / decoy / unmapped scaffolds (~0.1% of records) are excluded.
• Exclude stop-gain (alt = X) and same-AA records.

After filtering: 288,548 autosomal + 19,965 X + 30 Y + 135 MT = 308,678 missense variants with parsed chromosome.

2.2 Per-chromosome-class aggregation

Group variants into 4 classes: autosomes (chr1–22), X, Y, MT. Per class compute n_P, n_B, total, P_fraction, Wilson 95% CI.

3. Results

3.1 Per-chromosome-class Pathogenic fraction

The Wilson 95% CI on autosomes is [36.55%, 36.90%] (very tight at this N). The Wilson 95% CI on X is [47.15%, 48.54%] (tight at this N). The two CIs are non-overlapping by ~10 percentage points — the X-vs-autosomal P-fraction asymmetry is statistically robust.

3.2 The 11.12-percentage-point X-vs-autosomal asymmetry

The X-chromosome P-fraction (47.84%) exceeds the autosomal P-fraction (36.72%) by 11.12 percentage points — a ratio of 1.30× (47.84 / 36.72). In absolute terms, X-chromosome missense variants are 1.30× more likely to be classified Pathogenic than autosomal missense variants in our cache.

The N is large (19,965 X-chromosome variants; 288,548 autosomal). The asymmetry is highly statistically significant (chi-square p-value < 10⁻⁵⁰; not formally computed here but obvious from the non-overlapping Wilson CIs at this N).

3.3 The biological / curatorial mechanism

The X-Pathogenic asymmetry reflects multiple mechanisms:

• X-linked recessive disorders: classical examples DMD (Duchenne / Becker MD), F8 (hemophilia A), F9 (hemophilia B), G6PD (G6PD deficiency), CFTR-X (no, CFTR is autosomal), OTC (urea cycle disorder), GLA (Fabry disease), IDS (Hunter syndrome), WAS (Wiskott-Aldrich), HPRT1 (Lesch-Nyhan), RPGR (X-linked retinitis pigmentosa). Each of these is heavily clinically-curated.
• X-linked dominant disorders: MECP2 (Rett syndrome), PHEX (X-linked hypophosphatemia).
• X-linked intellectual disability: > 100 X-linked ID genes catalogued; a major contribution to ClinVar X-chromosome curation.
• Clinical research focus: X-linked disorders affect males more severely than females and are easier to ascertain through pedigree analysis; clinical research has historically focused intensively on the X chromosome.

The 47.84% X P-fraction reflects this concentrated curation focus.

3.4 The Y chromosome and mtDNA caveats

• Y chromosome (n=30): only 30 missense variants in our cache. The 70% Pathogenic fraction is based on small N and has wide CI. Y-chromosome genes are mostly testis-specific; few are Mendelian-disease-associated (SRY for sex determination; AZF region for spermatogenesis; KDM5D / RPS4Y1 for autoimmune / hematological associations). The high-N Y-chromosome data would require expanded clinical sequencing.
• mtDNA (n=135): 8.15% Pathogenic fraction. Mitochondrial variants are evaluated under different ACMG criteria (heteroplasmy, maternal inheritance) and the per-mtDNA submission pattern differs from nuclear DNA.

Both Y and MT classes are excluded from the autosomes-vs-X comparison because of the small N.

3.5 The autosomal baseline

The autosomal P-fraction of 36.72% is the corpus baseline for missense-Pathogenic-fraction in our cache. The cross-genome corpus average P-fraction (autosomes + X) is approximately (105,962 + 9,552) / (288,548 + 19,965) = 115,514 / 308,513 = 37.4%. The X-vs-autosomal-vs-corpus-baseline relationships are consistent: X is above the corpus baseline; autosomes are at the corpus baseline (autosomes account for >93% of variants).

4. Confound analysis

4.1 Stop-gain explicitly excluded

We filter alt = X. Reported numbers are missense-only.

4.2 X-chromosome-specific clinical-research focus

The X-chromosome P-fraction asymmetry is dominantly a clinical-research-focus signal, not a biological-pathogenicity signal. X-linked disorders have been intensively studied since the early 20th century; the per-chromosome P-fraction reflects this curation focus.

4.3 Per-gene composition differs

The X chromosome has approximately 800 protein-coding genes (chrX). Of these, a large fraction are well-curated X-linked disease genes (~100 X-linked intellectual disability genes alone). The X P-fraction asymmetry is partly driven by the higher fraction of disease-active genes per X-chromosome megabase.

4.4 Hemizygous males vs heterozygous females

X-linked recessive variants in males are hemizygous (1 allele expressing); the clinical sensitivity for detecting such variants is high. This contributes to the X-chromosome over-representation of Pathogenic submissions.

4.5 No correction for variant-type (within-missense)

We do not stratify the missense variants by substitution-class. The X-vs-autosomal P-fraction asymmetry could partially reflect different substitution-class distributions; this is an interesting follow-up but beyond scope.

4.6 Wilson CI assumes binomial sampling

Per-class counts are binomial. Wilson 95% CI is appropriate.

4.7 ClinVar curatorial bias

Pathogenic variants are over-reported in clinically-actionable disease genes; X-linked disease genes are heavily clinically-actionable. The X P-fraction reflects this curatorial bias.

5. Implications

1. X-chromosome missense variants are 1.30× more likely to be Pathogenic than autosomal variants in ClinVar (47.84% vs 36.72%; 11.12-percentage-point absolute difference; Wilson 95% CIs non-overlapping by ~10 pp).
2. The asymmetry is dominantly a clinical-research-focus signal on X-linked disorders (DMD, F8, OTC, GLA, IDS, WAS, RPGR, MECP2, PHEX, ~100 X-linked ID genes).
3. For variant-prioritization pipelines: a missense variant on the X chromosome carries a higher Pathogenic prior than on an autosome; per-chromosome-class prior is a useful metadata feature.
4. The Y-chromosome and mtDNA classes have small N and require expanded clinical sequencing for reliable per-class statistics.
5. The autosomal P-fraction of 36.72% is the per-chromosome-comparable corpus baseline.

6. Limitations

1. Stop-gain excluded (§4.1).
2. X-chromosome clinical-research focus drives the asymmetry (§4.2) — not pure biology.
3. Per-gene composition differs between X and autosomes (§4.3).
4. Hemizygous males vs heterozygous females ascertainment (§4.4).
5. No within-missense substitution-class stratification (§4.5).
6. ClinVar curatorial bias (§4.7).

7. Reproducibility

• Script: analyze.js (Node.js, ~30 LOC, zero deps).
• Inputs: ClinVar P + B JSON cache from MyVariant.info.
• Outputs: result.json with per-class counts, Pathogenic fractions, Wilson 95% CIs.
• Verification mode: 5 machine-checkable assertions: (a) Σ per-class counts = total; (b) all P-fractions in [0, 1]; (c) Wilson CIs contain the point estimates; (d) X P-fraction > autosomes P-fraction; (e) Wilson CIs of X and autosomes are non-overlapping.

node analyze.js
node analyze.js --verify

8. References

1. Landrum, M. J., et al. (2018). ClinVar. Nucleic Acids Res. 46, D1062–D1067.
2. Liu, X., Li, C., Mou, C., Dong, Y., & Tu, Y. (2020). dbNSFP v4. Genome Med. 12, 103.
3. Wu, C., et al. (2021). MyVariant.info. Bioinformatics 37, 4029–4031.
4. Wilson, E. B. (1927). Probable inference, the law of succession, and statistical inference. J. Am. Stat. Assoc. 22, 209–212.
5. Brown, L. D., Cai, T. T., & DasGupta, A. (2001). Interval estimation for a binomial proportion. Stat. Sci. 16, 101–133.
6. Bell, J., & Haldane, J. B. S. (1937). The linkage between the genes for colour-blindness and haemophilia in man. Proc. R. Soc. Lond. B 123, 119–150.
7. Lyon, M. F. (1961). Gene action in the X-chromosome of the mouse. Nature 190, 372–373. (X-inactivation reference.)
8. Ropers, H. H., & Hamel, B. C. J. (2005). X-linked mental retardation. Nat. Rev. Genet. 6, 46–57.
9. Karczewski, K. J., et al. (2020). gnomAD constraint spectrum. Nature 581, 434–443.
10. Richards, S., et al. (2015). ACMG/AMP variant interpretation guidelines. Genet. Med. 17, 405–424.