The Codon Memory Spectrum on the Double-Stranded Torus: Exact Zero-RNG Null Towers, Klein Four-Group Symmetry Breaking, the Two-Sided 16x16 Dinucleotide Screen, and the Algebra-to-Screen Bridge Across 54 RefSeq Genomes (Paper III Preprint, Autonomous Engine v2.4 & Complete 54-Taxon NCBI RefSeq CDS Archive)
Year: 2026
Type: preprint
Abstract: This record contains the complete manuscript (Paper III), LaTeX source, autonomous zero-RNG verification engine (codon_memory_spectrum_v2.py, v2.4), deterministic execution logs, JSON/Markdown data tables, 54-taxon biosphere figures, and the complete cached NCBI RefSeq coding-sequence (CDS) archive supporting the study: "The Codon Memory Spectrum on the Double-Stranded Torus: Exact Zero-RNG Null Towers, Klein Four-Group Symmetry Breaking, the Two-Sided 16×16 Dinucleotide Screen, and the Algebra-to-Screen Bridge Across 54 RefSeq Genomes" by Victor Logvinovich and Jean-Claude Perez (November 15, 2026). • Open-Source GitHub Repository: https://github.com/Dr-Logvinovich/codon-torus-syntax• Companion Paper I (v10): https://doi.org/10.5281/zenodo.23479667 (Concept DOI: https://doi.org/10.5281/zenodo.23287684)• Companion Paper II (v3): https://doi.org/10.5281/zenodo.23695321 (Concept DOI: https://doi.org/10.5281/zenodo.23508080) Building on Papers I and II—which established medium-range (~30-codon) conditional codon syntax and the double-stranded genetic-code fibration α_dup(c) = (α(c), α(ρ(c))) on the discrete torus T_4^3 = (Z/4Z)^3 (collapsing the single-strand synonymous dimension from 64 − 21 = 43 to 64 − 56 = 8)—this work constructs a complete, deterministic (zero-RNG) multi-lag theory of sequential memory across lags k = 1, ..., 60 codons. The autonomous engine is executed across 54 cellular NCBI RefSeq genomes (281,207 deduplicated genes, 141,272,088 sense codons, 423,816,264 nucleotides after strict frame, ACGT alphabet, internal-stop, and one-gene-one-vote longest-isoform filtering). FOUR PRINCIPAL MATHEMATICAL & EMPIRICAL RESULTS 1. Exact Zero-RNG Replacement of the Monte-Carlo Stack (Theorems A, B.1, C.2, D, H, K, N):Every stochastic control in the ten-null stack—H4 block permutations (closed-form O(n + q^2) bilinear expectation), H5 Markovian H2b surrogates (exact forward propagation with step-by-step mass invariant ||u_t||_1 = 1 to < 10^-12), H6 synonymous and H7 mirror multiset shuffles, and the four layer-shuffle towers Λ ∈ {het, aa, dup, mirror} (computed in 2F BLAS calls without clipping)—is replaced by an exact polynomial-time analytic expectation, verified against exhaustive permutation enumeration to 10^-15 – 10^-20. Along the alphabet quotient chains (codon_64 → mirror_32 → aa_20 and codon_64 → wobble_4 → gc3_2), coarse-graining monotonically expands the BIC-detectable Markov horizon p* across 54/54 taxa, while four eukaryotic genomes (Zebrafish, Xenopus tropicalis, Dictyostelium, Chlamydomonas) break the BIC barrier directly on the raw 64-codon and 20-AA alphabets. 2. Klein Four-Group (V_4) Symmetry Breaking & Commutator Invariants (Theorems M, L, Q, Q.2):The Klein four-group V_4 = {id, T, R, S} generated by time reversal T and antiparallel duplex inversion R splits pair space R^4096 into four orthogonal quadrants W_++ ⊕ W_+- ⊕ W_-+ ⊕ W_-- of exact dimensions 1056 ⊕ 1024 ⊕ 1024 ⊕ 992. Because every between-gene heterogeneity null E_het^(k) is time-symmetric (T E_het^(k) = E_het^(k)), its projection onto W_-+ ⊕ W_-- vanishes identically (< 10^-35 across all lags and all 54 taxa), establishing W_-+ ⊕ W_-- as a model-free witness of directed within-gene sequential order. Furthermore, the non-commutativity invariants ||[D_AA, R]||_F^2 = 943/36 ≈ 26.194444 and ||[Π_AA, R]||_HS^2 = 1962383/2592 ≈ 757.092207 act as a parameter-free detector of genetic-code reassignment, automatically isolating Candida albicans (Table 12, CTG: Leu → Ser; 11213/420 ≈ 26.697619 and 269863271/352800 ≈ 764.918569) from all 53 Table-1/11 genomes without species labels. 3. The Two-Sided 16×16 Dinucleotide Screen, Bias Lemmas & Central Methodological Correction (Theorems R, S, T, T.4, T.5, U, V, W, X; Lemmas 7.1–7.2):Order-2 nucleotide trigram models (H3b and phase-stratified H3b_phase) factor at every lag k through a rank-≤16 two-sided 16×16 screen Q_k = Q_1^k, seeing only the 225-df bridge I_bridge (225/3600 = 6.25% of sense degrees of freedom) while leaving 3375/3600 = 93.75% of sense degrees of freedom (D_5'lead = 675 df, D_3'wobble = 675 df, D_1↔3 = 2025 df) identically invisible. Under antiparallel duality (X, Y) ↦ (ρ(Y), ρ(X)), sense leading-base memory D_5'lead and antisense wobble memory D_3'wobble swap identically. The exact held-out decomposition d_3b(k) = Δ_x1 + D_suffix − D_gentogether with two finite-sample bias lemmas—proving that each screen layer carries a Miller–Madow floor df_l / (2 N_te) and that the plug-in generalisation gap scales as E[D_gen] ≃ (1 + r^-1) q(q − 1) / (2 N_te)—resolves the sign behaviour of held-out residuals:• Making the trigram null phase-aware removes 56.7%–89.9% of headline d_3b(k=1), which is shown by Theorem W to be frame-averaged stop-codon leakage (D_stop) and bridge KL misfit rather than sequential memory.• At k = 30, d_3b^phase becomes slightly negative in 10/54 small-cohort/extreme-composition prokaryotes where D_gen exceeds Δ_x1 + D_suffix^phase, exactly as predicted by Corollary 7.3.• Conversely, the model-free leading-base excess Δ_x1(k) = D_5'lead + D_1↔3 is strictly positive in 54/54 taxa at every reported lag including k = 30 (+0.002351 to +0.026840 nats/codon), and all 162 exact zero-RNG controls (Δ_5, Δ_6, Δ_7) have gene-level 95% confidence intervals bounded strictly above zero across 54/54 taxa. 4. The Algebra-to-Screen σ-Algebra Bridge & Topological Invariance (Theorems Y, Z, AA, AA.2):Both the duplex decomposition (361 + 779 + 2040 + 420 = 3600) and the dinucleotide screen (225 + 675 + 675 + 2025 = 3600) partition the same 3600 = 60^2 sense degrees of freedom. We prove that the right-codon prefix–AA incidence graph G_(O,A) is an acyclic forest on all eight NCBI translation tables (16 → 23 → 58 → 61 on Table 1/11 sense codons), whereas the left-codon suffix–AA graph G_(S,A) (16 → 57 → 61) has cycle rank (first Betti number) b_1 = 25, which remains an exact topological invariant under Table 12 codon reassignment (57 − 36 + 4 = 58 − 36 + 3 = 25). This yields the exact Kemeny–Snell Markovian lumpability split d_2b = I(S; A' | A) + I(X; A' | A, S) (703 + 76 = 779 df) and a self-dual 3×3 bi-filtration matrix satisfying M(B_anti) = M(B_sense)^T to ≤ 2.1 × 10^-17 nats/codon. ARCHIVE CONTENTS (65 FILES, 592.90 MB)• Codon_Memory_Spectrum_v2.pdf & Codon_Memory_Spectrum_v2.tex — Complete 30-page Paper III manuscript, proofs of Theorems A–AA, 54-taxon biosphere figures (Figs. 1–4), and full 54-taxon Appendix tables (Tables A1–A4).• codon_memory_spectrum_v2.py (v2.4) — Self-testing, zero-RNG Python verification engine implementing Theorems A–AA and exhaustive permutation self-tests (--self-test).• run_memory_v24_all_cached.log — Deterministic execution log across all 55 cached RefSeq assemblies (54 cellular taxa retained + SARS-CoV-2 11-CDS viral boundary control excluded by the ≥40-CDS guard; runtime 71.6 min).• results_memory_v24_full.zip — Complete output bundle containing memory_spectrum_v24.json, memory_spectrum_v24.md, and publication-grade 54-taxon figures (fig1_klein_quadrants.png, fig2_thmT_screen_layers.png, fig3_heldout_identity.png, fig4_tower_phase_boxes.png).• GCF_*_cds_from_genomic.fna.gz (55 archives) — Official NCBI RefSeq genomic CDS archives for all 54 cellular biosphere taxa (12 mammals including Human T2T-CHM13v2.0, 4 non-mammalian vertebrates, 6 invertebrates, 7 plants/green algae, 5 fungi, 5 protists, 13 bacteria, 2 archaea) plus the viral boundary control.• Companion forensic & reference caches — frontier_cds.zip, chimera_entrez_cache.zip, refseqgene_gb.zip, ncbi_live_viral.zip, and chr17_17q22_55154472_55254472.fasta for cross-paper reproducibility across Papers I–III. REPRODUCIBILITY1. Run deterministic startup self-tests (Theorems A–AA, all 8 NCBI tables, exhaustive permutation fixtures, 3-period de Bruijn B(4,6) null): python3 codon_memory_spectrum_v2.py --self-test2. Execute the full 54-taxon biosphere audit: python3 codon_memory_spectrum_v2.py --cds-dir . --out-dir results_memory_v24_full --all-cached --max-genes 6000 --max-lag 60 --kernel-mode backoff3. Source code, scripts (including replot_memory_v24.py), and updates are maintained at: https://github.com/Dr-Logvinovich/codon-torus-syntax
Authors: Logvinovich Victor, Perez, Jean claude
Institution: Brest State A.S. Pushkin University
Language: English
Subfield: Molecular Biology
Domain: Life Sciences
Open Access status: green
OpenAlex ID: W7233576271
Cites:
Cited by:
Related to:
