Seven Amino-Acid-Substitution Pairs Are Reachable Only Via CpG-Mutational-Pathway Source Codons (RC, RH, RL, RP, RQ, SW, TM): These 7 of 150 Single-Nucleotide-Reachable AA Pairs (4.67%) Are 2.92× Over-Represented in ClinVar Missense Variants (13.64% of 267,625 Variants) and Have a 24.08% Pathogenic-Fraction Vs 30.03% for Non-CpG-Only Pairs — Documenting the Genetic-Code-Architecture × CpG-Deamination Mutation-Rate Joint Quantification

Abstract

We enumerate the genetic-code-imposed CpG-mutational-pathway requirement for all 380 ordered amino-acid-substitution pairs (refAA, altAA). For each pair, we identify all (refAA-codon, altAA-codon) pairs differing in exactly one nucleotide position, and check whether the variant nucleotide position is part of a CG dinucleotide in the source codon (i.e., a CpG-context that mutates ~10× faster than non-CpG positions due to spontaneous deamination of 5-methylcytosine; Cooper & Krawczak 1990). Result: of 150 single-nucleotide-reachable AA-pairs:

• 7 pairs are CpG-required (reachable ONLY via CG-context source codons): RC, RH, RL, RP, RQ, SW, TM.
• 131 pairs are non-CpG-only (no CG-context paths).
• 12 pairs are mixed (both CpG and non-CpG paths).

The 7 CpG-required pairs comprise only 4.67% of reachable pairs, but account for 36,499 of 267,625 variants (13.64%) in ClinVar missense single-nucleotide variants (dbNSFP v4 (Liu et al. 2020) via MyVariant.info (Wu et al. 2021); stop-gain alt = X excluded). Over-representation ratio: 2.92× — CpG-required AA-pairs are observed in ClinVar at nearly 3× their expected frequency from genetic-code structure alone.

Mechanism: the 2.92× over-representation reflects the CpG-deamination mutation-rate amplification (Cooper & Krawczak 1990; Lynch 2010). Methylated cytosines in CG dinucleotides spontaneously deaminate to thymine at ~10× the background nucleotide-substitution rate, producing C>T transitions (and G>A on the opposite strand). The 7 CpG-required pairs (R↔C, R↔H, R↔L, R↔P, R↔Q, S↔W, T↔M — note the prevalence of arginine, encoded by CGN codons containing CpG) inherit this mutation-rate amplification. The 24.08% Pathogenic-fraction in CpG-required pairs (vs 30.03% in non-CpG-only) reflects the inverse relationship: the higher mutation rate produces more recurrent variants that accumulate as Benign in population databases, depressing the per-class Pathogenic-fraction. For variant-prioritization: a novel variant in a CpG-required pair has a 1.25× lower Pathogenicity prior (24.08% vs 30.03%) than a non-CpG-only variant — not because it's intrinsically less disease-causing, but because the higher mutation rate has accumulated more Benign instances in population data. Both classifications are non-circular (genetic-code-derived structural property of each AA pair) and provide actionable per-variant metadata.

1. Background

The CpG-deamination mutation rate amplification (Cooper & Krawczak 1990; Lynch 2010): methylated cytosines (5-methylcytosine) in CG dinucleotides spontaneously deaminate to thymine at ~10× the background nucleotide-substitution rate. The C>T transition (G>A on opposite strand) is the dominant CpG-mediated mutation. Approximately 70% of human point mutations causing genetic disease are at CpG sites (Cooper & Krawczak 1990).

The genetic code assigns codons to amino acids in a fixed mapping. Some amino-acid substitutions can be reached from a refAA codon to an altAA codon via a CG-containing source codon position; others can only be reached via non-CG source codons. The CpG-required AA pairs are those where the only single-nucleotide-mutation paths require a CpG-context source codon — these pairs inherit the mutation-rate amplification.

This paper enumerates the CpG-required AA-pair set, quantifies its empirical over-representation in ClinVar, and documents the per-class Pathogenicity-fraction asymmetry.

2. Method

2.1 Enumerate single-nucleotide-reachable pairs

For each of the 380 ordered (refAA, altAA) pairs with refAA ≠ altAA:

• Enumerate all codons encoding refAA and all codons encoding altAA.
• For each (refAA-codon, altAA-codon) pair with Hamming distance = 1, identify the changed position p (0, 1, or 2).
• Classify as CG-context if the codon position p is part of a CG dinucleotide within the codon (positions 0-1, 1-2, etc.).

2.2 Classify per-pair CpG dependency

A pair is CpG-required if all single-nucleotide-mutation paths require CG-context source codons. A pair is non-CpG-only if no path requires CG-context. A pair is mixed if both CG and non-CG paths exist.

2.3 ClinVar variant tabulation

For each ClinVar missense single-nucleotide variant (dbNSFP v4 via MyVariant.info; stop-gain alt = X excluded), classify into the per-pair CpG class. Tabulate Pathogenic and Benign counts. Compute Pathogenic-fraction with Wilson 95% CI (Brown et al. 2001).

2.4 Over-representation analysis

Compare per-class variant frequency (observed) to per-class pair frequency (expected from uniform per-pair distribution).

3. Results

3.1 The 7 CpG-required AA-pairs

The list is dominated by arginine-involving substitutions (5 of 7), because R is encoded by 6 CGN codons that contain CpG at positions 1-2. The remaining 2 (S→W, T→M) involve specific Ser-TCG and Thr-ACG codons that contain CpG at positions 2-3.

3.2 The 4-class distribution

The 7 CpG-required pairs are 2.92× over-represented in ClinVar variants relative to their share of reachable pair space (4.67% of pairs but 13.64% of variants). The 131 non-CpG-only pairs are slightly under-represented (87.33% of pairs but 73.71% of variants).

3.3 The Pathogenic-fraction asymmetry

The CpG-required pairs have lower P-fraction (24.08%) than non-CpG-only pairs (30.03%) by 5.95 percentage points — a 1.25× relative depression. Wilson 95% CIs are non-overlapping by ~5 pp.

3.4 The mechanism: mutation-rate vs Pathogenic-fraction inverse relationship

The CpG-required pairs have higher mutation rate (~10× CpG amplification) but lower Pathogenic-fraction:

• High mutation rate → more variants observed.
• More variants observed → more recurrent variants in healthy populations.
• Recurrent variants in healthy populations → more Benign curations in ClinVar.
• More Benign curations → lower per-class Pathogenic-fraction.

The 2.92× over-representation × 1.25× Pathogenicity-depression = the integrated effect of CpG-amplified mutation rate on ClinVar curation patterns. Both effects are consistent with the underlying CpG-deamination mechanism.

3.5 The Pathogenic count is still high in absolute terms

Despite the lower Pathogenic-fraction, the 7 CpG-required pairs account for 8,788 Pathogenic variants in ClinVar — substantial in absolute numbers due to the high mutation rate.

3.6 The 131 non-CpG-only pairs have higher P-fraction

Non-CpG-only AA pairs (e.g., A↔V, F↔Y, K↔N, etc.) have lower mutation rate, fewer recurrent variants, and a higher Pathogenic-fraction (30.03%) reflecting that observed variants are more often non-recurrent disease-relevant.

3.7 Implications for variant-prioritization

The CpG-required AA-pair classification is a non-circular metadata feature that captures mutation-rate effects on per-pair ClinVar curation:

• A novel variant in a CpG-required pair (RC, RH, RL, RP, RQ, SW, TM): prior P-fraction 24.08% — slightly Benign-leaning relative to non-CpG variants.
• A novel variant in a non-CpG-only pair: prior P-fraction 30.03% — slightly Pathogenic-leaning.

The 5.95-pp Pathogenic-fraction gap reflects the mutation-rate-driven curation asymmetry, not intrinsic biological severity. Variant-prioritization pipelines should incorporate this prior to avoid systematic mis-calibration on the 36,499 ClinVar variants in CpG-required AA pairs.

4. Confound analysis

4.1 Stop-gain explicitly excluded

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

4.2 The CpG-required classification is sequence-derived

The classification depends only on the genetic code structure and the AA-pair identity. Non-circular: independent of ClinVar curation, predictor scores, or any modern annotation.

4.3 The 10× mutation-rate amplification is well-established

CpG-deamination at ~10× background is documented in Cooper & Krawczak 1990, Lynch 2010, and many subsequent studies.

4.4 ClinVar curator labels are not gold-standard

Some labels are wrong. The reported Wilson 95% CIs reflect sampling variability.

4.5 The 7-pair CpG-required set is small

The 7 pairs are arginine-dominated. Arginine involvement creates a per-AA confound that's structurally inseparable from the CpG-required classification (arginine is the only AA with all 6 codons CG-containing).

4.6 The mixed-paths pairs (12 pairs) are intermediate

Mixed pairs (which have both CpG and non-CpG paths) have intermediate Pathogenic-fraction (25.84%) — between CpG-required (24.08%) and non-CpG-only (30.03%). Consistent with partial mutation-rate amplification.

4.7 The over-representation reflects observed variants, not biological frequency

The 2.92× over-representation in observed ClinVar variants reflects the integrated effect of mutation rate × ascertainment. The actual biological-event rate would be similar.

5. Implications

1. 7 of 150 single-nucleotide-reachable AA-pairs (4.67%) are CpG-required (reachable only via CG-context source codons): RC, RH, RL, RP, RQ, SW, TM.
2. These 7 pairs are 2.92× over-represented in ClinVar missense variants (13.64% of observed vs 4.67% of reachable pairs).
3. CpG-required pairs have a 24.08% Pathogenic-fraction vs 30.03% for non-CpG-only pairs — 1.25× relative depression.
4. The mechanism is CpG-deamination mutation-rate amplification (~10×): higher mutation rate produces more recurrent variants curated as Benign in population databases.
5. For variant-prioritization: the CpG-required AA-pair flag is a precomputable non-circular metadata feature with 5.95-pp Pathogenic-fraction depression that should be applied as a calibration correction.

6. Limitations

1. Stop-gain excluded (§4.1).
2. CpG-required classification is sequence-derived, non-circular (§4.2).
3. 10× CpG amplification is well-established (§4.3).
4. ClinVar labels not gold-standard (§4.4).
5. CpG-required set is small (7 pairs) and arginine-dominated (§4.5).
6. Mixed-paths pairs are intermediate (§4.6).
7. Over-representation reflects observed variants (§4.7).

7. Reproducibility

• Script: analyze.js (Node.js, ~70 LOC; embeds genetic code and CpG-context detection; zero deps).
• Inputs: ClinVar P + B JSON cache from MyVariant.info.
• Outputs: result.json with the 4-class structural counts, ClinVar variant counts per class, Wilson 95% CIs, and over-representation ratio.
• Verification mode: 5 machine-checkable assertions: (a) 7 CpG-required pairs identified; (b) over-representation ratio > 2.5×; (c) CpG-required P-fraction < 26%; (d) non-CpG-only P-fraction > 29%; (e) all 7 CpG-required pairs are R-involving except SW and TM.

node analyze.js
node analyze.js --verify

8. References

1. Cooper, D. N., & Krawczak, M. (1990). The mutational spectrum of single base-pair substitutions causing human genetic disease. Hum. Genet. 85, 55–74.
2. Lynch, M. (2010). Rate, molecular spectrum, and consequences of human mutation. Proc. Natl. Acad. Sci. USA 107, 961–968.
3. Crick, F. H. C. (1968). The origin of the genetic code. J. Mol. Biol. 38, 367–379.
4. Landrum, M. J., et al. (2018). ClinVar. Nucleic Acids Res. 46, D1062–D1067.
5. Liu, X., Li, C., Mou, C., Dong, Y., & Tu, Y. (2020). dbNSFP v4. Genome Med. 12, 103.
6. Wu, C., et al. (2021). MyVariant.info. Bioinformatics 37, 4029–4031.
7. Brown, L. D., Cai, T. T., & DasGupta, A. (2001). Interval estimation for a binomial proportion. Stat. Sci. 16, 101–133.
8. Bird, A. P. (1980). DNA methylation and the frequency of CpG in animal DNA. Nucleic Acids Res. 8, 1499–1504.
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.