From Form to Program: Medium-Range Conditional Codon Syntax Inside Real Genes — Strict Null-Model Evidence Across 54 Cellular RefSeq Taxa, a Codon-Mirror-Symmetry Control, and Two-Scale Phylogenetic Transfer (Preprint & Autonomous Pipeline v14.6)

This repository contains the updated preprint of the paper "From Form to Program: Medium-Range Conditional Codon Syntax Inside Real Genes — Strict Null-Model Evidence Across 54 Cellular RefSeq Taxa, a Codon-Mirror-Symmetry Control, Two-Scale Phylogenetic Transfer, and Intron–Exon Compartment Confinement" and the complete, self-testing source code for the autonomous verification suite (super_codon_autonomous_v14.py, version v14.5 audit-clean, and super_codon_proof_v14_6.py, version v14.6).

Building upon the "Form / Substance" framework introduced by J.-C. Perez (2010, 2013, 2015), this study operationalizes the distinction between static macroscopic genome invariants and dynamic sequential order in the language of held-out Markovian information theory. It mathematically separates the static compositional profile of coding DNA (form: marginal codon usage, isochore GC equilibria, folded dinucleotide/trigram frequencies, and single-stranded reverse-complement parity rules) from the conditional transition order P(c_{i+1} | c_i) inside true-frame coding sequences (substance / program).

Key Features & Scientific Contributions (v14.5 & v14.6 Expansion):

1. 54-Taxon Biosphere Scope + Viral Boundary Control (55 RefSeq Assemblies across 8 Full Audits):
Expands the audit from the primary 13-species reference panel to a curated 41-taxon extended biosphere panel, totaling 54 unique cellular taxa (eukaryotic nuclear and prokaryotic RefSeq assemblies spanning mammals, birds, amphibians, teleost fishes, insects, nematodes, land plants, green algae, fungi, protists, bacteria, cyanobacteria, and archaea) plus SARS-CoV-2 as an opt-in small-sample viral boundary control. The sampled biosphere spans the biological GC3 spectrum from extreme AT-rich (GC3 = 0.1416 in Dictyostelium discoideum) to extreme GC-rich (GC3 = 0.9177 in Thermus thermophilus).

2. Strict 10-Level Null-Model Hierarchy (H0–H7) with Formal 95% Bootstrap CIs:
Empirical codon-bigram models trained on 80% of deduplicated genes are evaluated on held-out 20% test genes (enforcing true reading frame, one-gene-one-vote isoform deduplication, start/stop codon trimming, family-aware gene splits, and per-record genetic code filtering unit-tested across NCBI translation tables 1, 2, 3, 4, 5, 6, 11, and 12) against ten null baselines: marginal codon usage (H0), boundary nucleotide context (H1), amino-acid bigram syntax with marginal synonymous usage (H2), amino-acid syntax with empirical codon-pair usage (H2b), folded dinucleotide composition (H3), folded order-2 nucleotide trigram dynamics (H3b), intra-gene block shuffles at L in {3, 10, 30, 100} codons (H4), amino-acid-preserving H2b conditional surrogates (H5), gene-specific synonymous multiset shuffles (H6), and reverse-complement mirror-pair shuffles (H7). All 52 cellular assemblies clearing the within-species sample-size guard (>= 20 units and >= 50,000 transitions in both train and test splits) pass the full STRICT stack at L = 30 codons (90 nt, matching the ribosomal exit tunnel scale) with strictly positive 95% bootstrap confidence intervals across all seven strict-stack quantities (d2b, d3, d3b, Delta_H4(30), Delta_5, Delta_6, Delta_7), while the 2 remaining small-split cellular assemblies (Leishmania major and Staphylococcus aureus) likewise exhibit strictly positive point estimates and 95% bootstrap lower bounds across all seven criteria.

3. Codon-Mirror-Symmetry Control (H7):
Permutes codons exclusively within each of the 32 unordered Watson–Crick reverse-complement mirror pairs {c, revcomp(c)} inside every held-out gene (extended Chargaff's second parity rule at the codon level, Perez 2013, Sec. 2.2) while destroying sequential order. Every cellular assembly passes H7 with a positive 95% bootstrap lower bound, demonstrating that conditional codon syntax is orthogonal to static strand-parity symmetry.

4. Two-Scale Cross-Species Phylogenetic Transfer:
- Coarse Scale (d2b_transfer > 0.02 + H4 gate): Uncovers a symmetric, vertebrate-wide conditional macro-layer (coarse_clean_0p02) across 20/20 directed pairs in the 5-taxon core vertebrate bridge, 20/20 directed pairs in a four-class vertebrate bridge (mammals, birds, amphibians, and two teleost orders), and 132/132 (100%) directed pairs across 12 mammalian species from six orders. Outside vertebrates, coarse transfer above 0.02 is localized mainly within narrower lower-rank clades (e.g., Caenorhabditis, core eudicots, monocot grasses, Enterobacteriaceae), with two high-GC prokaryotic exceptions into Pseudomonas aeruginosa (H. salinarum and M. tuberculosis) and one coarse high-GC plant-to-alga transfer (Zea mays -> Chlamydomonas reinhardtii).
- Fine Scale (Target-Scored H5 Conservation & Primate–Rodent Directional Asymmetry): Using a target-scored H5 surrogate metric (gap = Delta_5_transfer_target - Delta_5_within_target, ratio = Delta_5_transfer / Delta_5_within) across a focal set of 34 mammalian contrasts, resolves bidirectional conservation within Primates (Human <-> Chimp, Gorilla, Macaque), within Rodents (Mouse <-> Rat), and within Ungulates (Cow <-> Pig), alongside a reproducible Euarchontoglires directional asymmetry: examined focal primate-to-rodent transfers exhibit small negative gaps (gap in [-0.0065, -0.0032]), whereas every examined rodent-to-primate transfer exhibits a large positive overshoot (gap in [+0.0171, +0.0207]) despite near-zero GC3 distance (Delta_GC3 <= 0.011 across core and focal contrasts; <= 0.008 within the 12-mammal audit).

5. Compartment-Level Intron–Exon Confinement, Splice-Junction Phase & TP53 Splicing Trajectory (v14.6):
An autonomous compartment-level control across 12 human NCBI RefSeqGene loci (TP53, TP63, TP73, BRCA1, PTEN, RB1, GAPDH, ACTB, HBB, KRAS, MYC, and APOE) evaluates four structural states per locus (mature concatenated exons, whole unspliced pre-mRNA in frame 0, whole unspliced pre-mRNA across all 3 reading frames, and excised introns) against the human T2T-CHM13v2.0 reference transition model. Within the exact same genomic locus, the conditional syntax residual d3b is positive in mature concatenated exons (panel mean +0.1721 nats/codon; 0.00 internal stops/100 codons) and negative in both excised introns (mean -0.3210 nats/codon; mean exon-minus-intron contrast +0.49305 nats/codon in 12/12 genes) and whole unspliced pre-mRNA across all three reading frames (mean exon-minus-best-3-frame contrast +0.42152 nats/codon in 12/12 genes), while 32-pair mirror asymmetry (Mirror_L1) is higher in spliced exons than in introns in 12/12 loci (mean 0.4109 vs. 0.1612). These results support an operational compression-operator interpretation of splicing: intronic and unspliced compartments act as relatively higher-entropy, more mirror-symmetric compositional buffers (form), whereas exon concatenation enriches a symmetry-broken, more time-directed coding program (with the intronic baseline and sequence-level geometric invariants completely independent of CDS training data, and single-locus overlap representing < 10^-3 of the pooled reference transitions).

6. Compositional Hardware Guard & Directional Small-Sample Guard:
Explicitly separates true grammatical divergence from extreme compositional mismatches (GC/aa artifact candidates when |Delta_GC3| > 0.12 or |Delta_acid| > 0.08, such as halophilic adaptation in H. salinarum) and prevents small-sample distortion (such as the 11-ORF SARS-CoV-2 genome) via automated directional sample-size guards.

Included Files:
- code_logos_perez_v14_6.tex.pdf — The full text of the updated preprint in PDF format (24 pages, including Main Tables 1–10 and Appendices A–E with Tables 11–14).
- code_logos_perez_v14_6.tex — The complete LaTeX source code of the manuscript.
- super_codon_autonomous_v14.py — The self-testing autonomous Python verification pipeline (v14.5 audit-clean) for the 55-assembly multi-species audits, featuring built-in Stage 00 unit tests (--self-test) for NCBI translation tables 1, 2, 3, 4, 5, 6, 11, and 12 and 32 reverse-complement mirror pairs.
- super_codon_proof_v14_6.py — The extended autonomous Python verification pipeline (v14.6) adding the NCBI E-utilities RefSeqGene GenBank parser (join(...) splicing reconstruction unit-tested in --self-test), 4-state Intron–Exon compartment audits, Perez 32-mirror-pair L1 asymmetry and Markov flow irreversibility metrics, splice-junction phase reconstruction audits, and in-silico 5'->3' progressive splicing trajectories (--p53-audit, --intron-exon-audit).
- logs.zip — Complete execution logs, JSON/Markdown reports, 95% bootstrap confidence interval tables, exhaustive coarse transfer matrices, focal H5-transfer conservation reports (runs/ for all eight v14.5 audits reproduced with seed=20261014), and extended 12-gene RefSeqGene compartment & splicing audit logs (results_v14_6/: p53_extended_audit.log, p53_intron_exon_audit.json, p53_intron_exon_audit.md reproduced with seed=20261011).
- GCF_*_cds_from_genomic.fna.gz — Cached NCBI RefSeq coding-sequence archives for offline reproducibility.

Attribution & Scope Note:
This verification suite measures the dynamical/program layer (conditional codon transition syntax inside annotated coding sequences and spliced vs. unspliced RefSeqGene compartments) and treats static compositional and strand-mirror symmetries as form absorbed by H0, H1, H3, H3b, and H7. Whole-chromosome non-coding static attractors (such as the Human Genome Optimum / LOH ratios) concern zero-order marginal statics on different genomic objects and remain a separate, orthogonal measurement within the broader form/substance research programme.