Per-Gene Stop-Gain-vs-Missense Pathogenicity Gap Identifies a 793-Gene Loss-of-Function-Dominant Disease Subset (63.80% of 1,243 ClinVar-Eligible Genes With ≥20 Missense AND ≥10 Stop-Gain Variants): Mean Per-Gene Stop-Gain Pathogenic-Fraction Is 98.67% Vs Missense 38.67% — A 60-pp Gap, With 249 Genes (20.03%) Showing ≥90 pp Gap Including CHAMP1, AMER1, MYO18B, CEP250, NRXN1, LAMC3, PCLO All at 0% Missense Vs 100% Stop-Gain Pathogenic

Abstract

We compute the per-gene Pathogenic-fraction gap between stop-gain and missense variants for ClinVar (Landrum et al. 2018) variants in dbNSFP v4 (Liu et al. 2020) via MyVariant.info (Wu et al. 2021). Stop-gain variants identified as aa.alt = X and aa.ref ≠ X; missense identified as aa.alt ≠ X and aa.ref ≠ X and aa.ref ≠ aa.alt. Restricted to 1,243 genes with ≥ 20 missense AND ≥ 10 stop-gain variants. Result:

• Mean per-gene missense Pathogenic-fraction: 38.67%.
• Mean per-gene stop-gain Pathogenic-fraction: 98.67%.
• Mean per-gene gap (stop-gain − missense): 60.00 percentage points.
• Genes with stop-gain − missense gap ≥ 50 pp (LoF-dominant): 793 (63.80%).
• Genes with stop-gain − missense gap ≥ 90 pp (extreme-LoF-dominant): 249 (20.03%).

The top 20+ extreme-LoF-dominant genes have 0.0% missense Pathogenic AND 100.0% stop-gain Pathogenic: CHAMP1, AMER1, COL9A1, ANO6, WAC, MYO18B, ARCN1, SCAF4, CEP250, CENPF, C6, DSG1, NRXN1, LAMC3, C8A, CEP104, PHF21A, TANC2, ADCY10, PCLO, TECPR2, DNAAF2, CNTN2, KIDINS220, plus PCNT at 0.5% / 100.0% (gap 99.5 pp). Mechanism: these are loss-of-function-dominant Mendelian disease genes where the disease mechanism is gene knockout / haploinsufficiency — truncating variants abolish protein function and cause disease (curator-Pathogenic at near-100%), while typical missense substitutions retain partial function and are tolerated (curator-Benign in our sample). The biological interpretation: for these genes, only loss-of-function variants are pathogenic; partial-function missense variants do not produce phenotype. For variant-prioritization: in the 793 LoF-dominant genes, a missense variant of unknown significance has a strong Benign prior (~5-30%) while a stop-gain variant has a strong Pathogenic prior (~95-100%). The per-gene stop-gain-vs-missense gap is a precomputable LoF-dominance-classifier that distinguishes disease-mechanism types and informs variant-class-specific priors.

1. Background

Mendelian disease genes can be classified by their disease mechanism:

• Loss-of-function (LoF) dominant: gene knockout / haploinsufficiency causes disease. Truncating variants (nonsense, frameshift, splice-disrupting) are highly Pathogenic; partial-function missense variants are typically tolerated. Examples: many tumor suppressor genes, transcription factors with dosage sensitivity, hemizygous-male X-linked genes.
• Missense-dominant: specific gain-of-function or dominant-negative missense substitutions cause disease. Truncating variants may be tolerated (because of haploinsufficiency tolerance). Examples: oncogene activating mutations (RAS family), dominant-negative missense in some receptors.
• Mixed: both classes contribute to disease.

The per-gene stop-gain-vs-missense Pathogenic-fraction gap is a quantitative classifier of LoF-dominance: large positive gaps indicate LoF-dominant; near-zero gaps indicate similar Pathogenicity for both variant classes.

This paper measures the per-gene gap distribution across 1,243 ClinVar-eligible genes and identifies the LoF-dominant subset.

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 dbnsfp.aa.ref, dbnsfp.aa.alt, dbnsfp.genename.

2.2 Variant-class classification

• Stop-gain (nonsense): aa.alt = X AND aa.ref ≠ X.
• Missense: aa.alt ≠ X AND aa.ref ≠ X AND aa.ref ≠ aa.alt.
• Same-AA records excluded.

2.3 Per-gene per-class tabulation

Per gene: count msP, msB (missense Pathogenic, Benign) and sgP, sgB (stop-gain Pathogenic, Benign).

2.4 Per-gene gap

Per-gene gap = (stop-gain Pathogenic-fraction) − (missense Pathogenic-fraction).

2.5 Eligibility threshold

Restrict to genes with ≥ 20 missense AND ≥ 10 stop-gain variants for stable per-gene fraction estimation.

After filtering: 1,243 genes retained.

3. Results

3.1 Per-gene aggregate statistics

The ~60-pp average gap reflects the dominant pattern: stop-gain variants are nearly always Pathogenic (~99% on average), while missense variants are Pathogenic at ~39% on average (close to the global ~28% rate).

3.2 The LoF-dominant gene subsets

63.80% of analyzed genes have stop-gain − missense gap ≥ 50 pp — clear LoF-dominance signature. 20.03% have ≥ 90 pp gap — extreme LoF-dominance.

3.3 The extreme-LoF-dominant subset (top 25)

The top 25 extreme-LoF-dominant genes are dominated by:

• Tumor suppressors (AMER1).
• Centrosome / cilia genes (CEP250, CEP104, CENPF, DNAAF2 — primary ciliopathies).
• Multi-domain scaffold proteins (NRXN1 neurexin, LAMC3 laminin, PCLO piccolo, PCNT pericentrin).
• Ion channels / membrane (ANO6, CNTN2).
• Complement components (C6, C8A — complement deficiencies are recessive LoF).
• Cytoskeletal / structural (COL9A1 collagen, MYO18B myosin, DSG1 desmoglein).

These genes share the haploinsufficiency / homozygous-LoF disease mechanism where partial-function missense variants do not produce phenotype.

3.4 The smallest-gap genes

The smallest-gap genes (gap < 5 pp) include:

• PCSK9: ms 30.7% > sg 5.9% (gap −24.8 pp). PCSK9 is a gain-of-function gene where missense activate the protein causing hypercholesterolemia; truncating variants are LoF and protective for cardiovascular disease (curated as Benign or even protective).
• GABRD: ms 2.5% > sg 0.0%. Receptor with mostly tolerated variants.
• ALPL: ms 95.0% ≈ sg 94.9%. Hypophosphatasia — missense and stop-gain similarly Pathogenic.
• SMN1: ms 100% = sg 100%. Spinal muscular atrophy — both classes Pathogenic.
• PAH, GCK, GBA, RPE65: missense and stop-gain similarly Pathogenic (recessive disorders where any LoF variant suffices).

These genes are non-LoF-dominant in different ways: PCSK9 is gain-of-function-dominant; PAH/GCK/GBA are recessive-LoF-dominant where even missense are LoF.

3.5 The mechanism: haploinsufficiency vs gain-of-function classification

The per-gene gap is a quantitative classifier of disease mechanism:

• Large gap (LoF-dominant haploinsufficiency): gene knockout produces phenotype; partial-function missense are tolerated.
• Near-zero gap (recessive-LoF or constitutive disease): any LoF variant (missense or stop-gain) produces phenotype.
• Negative gap (gain-of-function-dominant): missense activate the protein; truncating variants are protective.

The 793-gene LoF-dominant subset (63.80%) and 249-gene extreme-LoF subset (20.03%) provide concrete classifications.

3.6 Implications for variant-prioritization

For variant-prioritization in LoF-dominant genes:

• A novel stop-gain variant: prior P-fraction ~95-100%. Pathogenic with high confidence.
• A novel missense variant of unknown significance: prior P-fraction ~5-30% (LoF-dominant gene baseline). Benign-leaning, requires manual review.

The per-gene gap is a precomputable disease-mechanism classifier from ClinVar metadata, providing variant-class-specific priors.

4. Confound analysis

4.1 Variant-class identification by aa.alt = X

Stop-gain variants are aa.alt = X; missense are non-X. Standard dbNSFP convention.

4.2 The ≥ 20 missense + ≥ 10 stop-gain threshold

Genes with sparse data are excluded. The 1,243 retained genes represent the well-curated subset.

4.3 ClinVar curator labels are not gold-standard

Some labels are wrong. The reported per-gene rates reflect curator-assigned data.

4.4 The 0% missense Pathogenic in many LoF-dominant genes is a sample-size effect

Some of the 0% missense Pathogenic rates may reflect that no Pathogenic missense have been submitted to ClinVar yet, not that none exist. Future curation may identify some.

4.5 The 100% stop-gain Pathogenic in many genes assumes NMD-triggering

C-terminal NMD-escaping stop-gains may be tolerated in some cases. The per-gene 100% sg-Pathogenic includes any tolerable C-terminal cases as small-N noise.

4.6 The disease-association annotations are post-hoc

Gene-disease links cited in §3.3 are from OMIM / GeneReviews lookup, post-hoc to the analysis.

4.7 The PCSK9 gain-of-function pattern is well-known

The PCSK9 gap inversion is a well-documented case (Cohen et al. 2006). Other gain-of-function genes may have similar patterns.

5. Implications

1. Per-gene stop-gain-vs-missense Pathogenicity gap is 60.00 pp on average across 1,243 ClinVar-eligible genes (mean ms 38.67%, mean sg 98.67%).
2. 63.80% of genes (793) have gap ≥ 50 pp — clear LoF-dominance signature.
3. 20.03% of genes (249) have gap ≥ 90 pp — extreme-LoF-dominance with 0% ms / 100% sg pattern.
4. The mechanism is haploinsufficiency / homozygous-LoF: partial-function missense are tolerated; truncating LoF variants cause disease.
5. For variant-prioritization in LoF-dominant genes: missense variants carry a low Pathogenic prior (~5-30%); stop-gain variants carry a high prior (~95-100%) — variant-class-specific priors should be applied.

6. Limitations

1. Stop-gain identified by aa.alt = X standard convention (§4.1).
2. ≥ 20 ms + ≥ 10 sg threshold restricts to 1,243 well-curated genes (§4.2).
3. ClinVar labels not gold-standard (§4.3).
4. 0% ms Pathogenic may reflect curation gap rather than true tolerance (§4.4).
5. NMD-escape C-terminal stop-gains may be tolerated (§4.5).
6. Disease annotations are post-hoc (§4.6).
7. Gain-of-function genes (PCSK9) are exceptions to the LoF-dominance pattern (§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-gene ms/sg counts, P-fractions, gap, and the top-30 LoF-dominant gene list.
• Verification mode: 5 machine-checkable assertions: (a) mean ms < 45%; (b) mean sg > 95%; (c) ≥ 700 genes with gap ≥ 50 pp; (d) ≥ 200 genes with gap ≥ 90 pp; (e) total eligible genes > 1,000.

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. Cohen, J. C., Boerwinkle, E., Mosley, T. H., & Hobbs, H. H. (2006). Sequence variations in PCSK9, low LDL, and protection against coronary heart disease. N. Engl. J. Med. 354, 1264–1272.
5. Karczewski, K. J., et al. (2020). gnomAD constraint spectrum. Nature 581, 434–443.
6. Cassa, C. A., et al. (2017). Estimating the selective effects of heterozygous protein-truncating variants from human exome data. Nat. Genet. 49, 806–810.
7. MacArthur, D. G., et al. (2012). A systematic survey of loss-of-function variants in human protein-coding genes. Science 335, 823–828.
8. Wright, C. F., et al. (2019). Paediatric genomics: diagnosing rare disease in children. Nat. Rev. Genet. 19, 253–268.
9. Adam, M. P., et al. (2022). GeneReviews. University of Washington, Seattle.
10. Richards, S., et al. (2015). ACMG/AMP variant interpretation guidelines. Genet. Med. 17, 405–424.