Per-Reference-Amino-Acid Benign Variant Enrichment in ClinVar Missense Records: Arginine Tops the List at 2.92× Vs Human Proteome Composition (16.4% of All Benign Missense Despite Being Only 5.6% of the Proteome) Across 191,030 Benign Records — Driven by CpG-Hotspot Mutational Bias

Abstract

We measure the per-reference-amino-acid Benign-variant enrichment in ClinVar missense single-nucleotide variants (Landrum et al. 2018), restricted to the Benign subset (alt ≠ X excluded; dbNSFP v4 (Liu et al. 2020) annotation via MyVariant.info (Wu et al. 2021)). The metric: per-AA Benign-share (n_per-AA / total Benign) divided by the human proteome AA composition baseline (UniProt SwissProt 2023). Result: across 191,030 Benign missense records, per-AA Benign enrichment spans an 11× range from 0.27× (Trp) to 2.92× (Arg): R 2.92× (16.43% of Benign missense / 5.62% of proteome); M 1.45×; A 1.42×; V 1.38×; P 1.29×; T 1.26×; I 1.17×; N 1.06×; G 0.99×; H 0.92×; D 0.89×; S 0.82×; E 0.67×; Q 0.58×; K 0.50×; C 0.46×; Y 0.45×; L 0.38×; F 0.36×; W 0.27×. The interpretation is consistent with mutational rates rather than selection: Arg is over-represented in Benign because of the well-known CpG-hotspot mechanism (CpG dinucleotides at the CGN arginine codons mutate at ~10× the background rate; Cooper & Krawczak 1990), producing many R-derived substitutions in tolerant positions. Trp, Phe, Leu, Tyr, Cys are under-represented in Benign because they are functionally constrained and rarely tolerate substitution: any substitution disrupts hydrophobic-core packing or aromatic stacking with high probability of being Pathogenic. The complementary observation: the per-AA Pathogenic-vs-Benign asymmetry mirrors this Benign distribution inversely — Trp is over-represented in Pathogenic (5.3× per independent analyses) because Benign substitutions of Trp are rare. For variant-prioritization pipelines: per-AA Benign-enrichment (or its inverse) is a useful baseline-frequency-correction prior. A novel R missense variant has a high-prior probability of being Benign; a novel W missense variant has a high-prior probability of being Pathogenic, by reference-AA identity alone.

1. Background

The per-reference-amino-acid distribution of ClinVar Pathogenic and Benign missense variants is shaped by two distinct factors:

• Mutation rate: codon-level mutation rates differ across amino acids; CpG dinucleotides have ~10× elevated mutation rate (Cooper & Krawczak 1990; Lynch 2010). Amino acids with CpG-rich codons (Arg, with 6 codons including CGN) accumulate variants at higher rates.
• Selection: functionally constrained positions are less tolerant of substitution; these positions accumulate Pathogenic variants disproportionately.

Pathogenic-variant enrichment captures the combined effect of mutation rate × selection. Benign-variant enrichment captures predominantly the mutation rate (since Benign variants by definition pass selection, the per-AA Benign enrichment reflects how often the AA gets mutated, weighted by selection-tolerance).

This paper measures per-AA Benign enrichment as a measurement of mutational-rate-driven baseline variant frequency.

2. Method

ClinVar Benign missense (alt ≠ X) variants from MyVariant.info / dbNSFP v4. For each variant: extract dbnsfp.aa.ref (first if array). Group by ref AA. Compute per-AA Benign-share = n_AA / total_Benign. Compute Benign-enrichment = per-AA-Benign-share / per-AA-proteome-share (using UniProt SwissProt 2023 reference proteome composition).

Wilson 95% CI on the per-AA share (Wilson 1927; Brown et al. 2001).

3. Results

3.1 Per-reference-AA Benign enrichment (sorted descending)

The 20 reference AAs span an 11× Benign-enrichment range (2.92 / 0.27).

3.2 The Arg over-representation (2.92×)

Arginine accounts for 16.4% of all Benign missense variants in ClinVar despite being only 5.6% of the human proteome — a 2.92× enrichment. Mechanism:

• Arg has 6 codons (CGT, CGC, CGA, CGG, AGA, AGG); the most-degenerate among basic AAs.
• The CGN subset (4 codons) is CpG-dinucleotide-containing.
• Methylated cytosines at CpG dinucleotides deaminate to thymines at ~10× the background rate (Cooper & Krawczak 1990; Lynch 2010), producing CGN → TGN (Arg → Cys) and CGN → CAN (Arg → Gln) substitutions at elevated rates.
• These mutations occur frequently across the genome — including in functionally tolerant positions — populating the Benign category disproportionately.

The Arg Benign enrichment of 2.92× is the largest per-AA Benign-share observation; it reflects the CpG-hotspot mechanism more than any selection signal.

3.3 The Trp under-representation (0.27×)

Tryptophan accounts for only 0.36% of Benign missense variants despite being 1.31% of the proteome — a 0.27× under-representation. Mechanism:

• Trp has 1 codon (TGG); the lowest mutation-opportunity per residue.
• Trp residues are functionally constrained (largest amino acid; unique indole aromatic ring; participates in aromatic-aromatic stacking, π-cation interactions, "Trp belt" membrane-interface clustering).
• Most Trp substitutions are functionally consequential and therefore classified Pathogenic, not Benign.

The Trp Benign-fraction at 0.36% is the lowest per-AA value observed; the corresponding Pathogenic enrichment for Trp (per independent analyses) is 5.3× — the inverse of this pattern.

3.4 The chemistry-class clustering

The Benign-enrichment values cluster into two regimes:

Tier 1 — Over-represented in Benign (enrichment > 1.0): R, M, A, V, P, T, I, N, G — 9 AAs. These are predominantly the small or moderate hydrophobic / polar AAs whose substitutions tend to be functionally tolerable.

Tier 2 — Under-represented in Benign (enrichment < 1.0): H, D, S, E, Q, K, C, Y, L, F, W — 11 AAs. These are predominantly the structurally-constrained (Cys disulfide; aromatic Trp/Phe/Tyr; charged D/E/K) or large hydrophobic (Leu) AAs whose substitutions are more often functionally consequential.

The boundary is approximately at the proteome-baseline 1.0× enrichment, with G and H sitting near the boundary.

3.5 Comparison to per-AA Pathogenic enrichment

Independent per-AA Pathogenic-enrichment analyses (companion-internal computations on the same dataset) yield approximately the inverse pattern: Trp at 5.3× Pathogenic enrichment, Arg at 2.8× Pathogenic enrichment (both elevated; Arg from CpG mutation rate, Trp from selection).

The cross-axis structure: Arg is enriched in BOTH Pathogenic and Benign (consistent with high mutation rate); Trp is enriched ONLY in Pathogenic (consistent with selection without high mutation rate); Leu/Phe are under-represented in BOTH (consistent with low per-residue mutation rate AND functional constraint).

4. Confound analysis

4.1 Stop-gain explicitly excluded

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

4.2 ClinVar curatorial bias

Benign variants are over-represented in population-genome-derived submissions (gnomAD-derived, large sequencing studies). The per-AA Benign enrichment reflects the AA composition of population variation, which is well-correlated with mutational opportunity but also includes a small selection component.

4.3 Codon-mutability is the dominant signal

The per-AA Benign enrichment is dominated by codon-mutability rather than selection. Arg's 2.92× enrichment is mostly the CpG-hotspot rate; Trp's 0.27× under-representation is partly the single-codon constraint (low opportunity) and partly the functional-constraint selection.

4.4 Per-isoform first-element AA

We use the first finite element of dbnsfp.aa.ref. ~5% per-isoform mismatch.

4.5 Proteome baseline is reviewed-SwissProt-only

The proteome composition baseline (UniProt 2023) is computed over reviewed Swiss-Prot human entries (~20,000 proteins). TrEMBL-annotated entries differ slightly in composition.

4.6 Wilson CI assumes binomial sampling

Per-AA counts are binomial. Wilson 95% CI is appropriate (Brown et al. 2001).

5. Implications

1. Per-reference-AA Benign-variant enrichment spans 11× across the 20 standard AAs in ClinVar missense, from 0.27× (Trp) to 2.92× (Arg).
2. Arg is the most Benign-enriched ref AA (16.4% of Benign missense; 2.92× over proteome) — driven by CpG-hotspot mutation rate.
3. Trp is the most Benign-depleted ref AA (0.36% of Benign missense; 0.27× under proteome) — single codon + functional constraint.
4. The chemistry tier-split (Tier 1 over-represented: R, M, A, V, P, T, I, N, G; Tier 2 under-represented: H, D, S, E, Q, K, C, Y, L, F, W) approximately corresponds to small/flexible vs structurally-constrained AAs.
5. For variant-prioritization pipelines: per-AA Benign-enrichment is a useful baseline-frequency-correction prior; novel R variants are more likely Benign; novel W variants are more likely Pathogenic, by reference-AA identity alone.

6. Limitations

1. Stop-gain excluded (§4.1).
2. ClinVar Benign curatorial bias (§4.2) — population-genome-derived dominant.
3. Codon-mutability dominates (§4.3) — the signal is mostly mutation rate, not pure selection.
4. Per-isoform first-element AA (§4.4).
5. Proteome baseline is SwissProt-only (§4.5).

7. Reproducibility

• Script: analyze.js (Node.js, ~40 LOC, zero deps).
• Inputs: ClinVar Benign JSON cache from MyVariant.info; UniProt SwissProt 2023 proteome AA composition (hard-coded).
• Outputs: result.json with per-AA Benign counts, Benign shares, proteome shares, enrichment factors.
• Verification mode: 5 machine-checkable assertions: (a) Σ proteome AA percentages ≈ 100; (b) Σ per-AA Benign counts = total Benign; (c) Arg has the highest enrichment; (d) Trp has the lowest; (e) Wilson CIs contain the point estimates.

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. Cooper, D. N., & Krawczak, M. (1990). The mutational spectrum of single base-pair substitutions causing human genetic disease. Hum. Genet. 85, 55–74.
7. Lynch, M. (2010). Rate, molecular spectrum, and consequences of human mutation. PNAS 107, 961–968.
8. The UniProt Consortium (2023). UniProt: the Universal Protein Knowledgebase in 2023. Nucleic Acids Res. 51, D523–D531.
9. Karczewski, K. J., et al. (2020). gnomAD constraint spectrum quantified from variation in 141,456 humans. Nature 581, 434–443. (Population-variation-derived Benign reference.)
10. Akashi, H., & Gojobori, T. (2002). Metabolic efficiency and amino acid composition in the proteomes of Escherichia coli and Bacillus subtilis. PNAS 99, 3695–3700. (Trp metabolic-cost reference.)