Unified deterministic interpreter

One deterministic interpreter reads a locus in five layers: the material γ, its R19 switch, its A4 coordinate (shell, anchor, loops, anchor-relative phase), the methylation regime read the way each clade writes it, and the runtime quantities it flags but never assigns. The coordinate grammar is restored; the global helical claim is corrected to an anchor-relative contact read.

Earlier chapters split the read in two: the A4 grammar gave structure and position, while the cross-kingdom and clade chapters measured γ and methylation as global statistics without the coordinate. This chapter unites them into a single engine that, for one locus, reads the material γ (immutable: human_SOX2 = 1.287315), the R19 double-well, the A4 coordinate — which shell, the nearest anchor and its strength, the loops built from real motors parsed from re-acquired feature tables, and the anchor-relative helical phase — and the methylation regime from the clade-matched reader, flagging the runtime sign, fill, and role rather than assigning them. Two accuracy deltas are closed: the structural coordinate is restored for every locus, and the prior global periodicity claim is corrected to the anchor-relative contact read, with the old surrogate retired. The methylation layer is reproduced bit-for-bit and promoted over raw CpG density, which conflates with GC.

Read every layer of one locus with one deterministic engine

LCT promoter: γ 1.315 · mid-shell · anchor 1065 bp · 1 loop / 5 motors · face 0.44 → contact False · CpG O/E 0.335 (raw density 0.0148) → VERTEBRATE_global_CG. Plant + insect regions read coordinate and regime together; methylation reproduces §12 across 52 quantities; negative control O/E 0.335 → 0.998.

The structure-and-position grammar and the clade-matched methylation reader are not two papers but two layers of one read. Restoring the coordinate makes the headline lactase locus report its shell, anchor, loops, and helical face; correcting the helical claim replaces a global periodicity surrogate with the physically meaningful phase between an element and its anchor; keeping methylation as O/E removes the GC confound that raw density carries.

Grade: principle-demonstration (mechanical Layer-1 read; sign, fill, and role stay Layer-2)

Canonical reproduction → repro/dna/13-unified-deterministic-interpreter/

One locus, five layers, one engine

The interpreter takes a wide genomic window, an element offset, and the element sequence, and returns five reads that until now lived in different modules. The material γ and its R19 switch come from the locked grammar; the coordinate comes from the A4 pipeline; the methylation regime comes from the clade-matched reader; the runtime quantities are flagged. The locked grammar and the methylation engine are imported as a single source and pinned by hash, so nothing in the earlier chapters is rewritten.

Every read is sequence-derived and deterministic, and two independent runs are byte-identical. What the sequence does not license — the brake-or-accelerator sign, the runtime fill, the cascade role — is returned as an explicit Layer-2 flag, never assigned from the sequence.

The coordinate grammar, restored

The cross-kingdom and clade chapters measured γ and methylation as global statistics but did not read where an element sits. This chapter restores that read: for the lactase promoter it reports a mid-shell position, a nearest anchor 1065 bp away, one co-anchored loop among five motors in the window, and an anchor-relative helical face of 0.44, which is not contact-competent. The loops are built from real motors parsed from feature tables re-acquired for each frozen region and stored with their provenance.

Where a window is gene-poor the read says so. The human chromosome-1 region carries only two motors at its edge, so its loop count is marked open and names the dataset that would close it — the full annotation for that region. The structure is read where the annotation supports it and flagged where it does not.

The helical claim, corrected

An earlier claim reported that a global periodicity in the A/T signal sits above a composition-matched shuffle in every genome. That is a composition surrogate for nucleosome spacing read from an arbitrary window edge, not a statement about whether a specific element can touch a specific partner. It is retired as a structural claim about an element.

The correct read is the phase between two loci. Using the locked rise of 3.4 Å and twist of 34.29° per base pair, the engine asks whether an element and its nearest anchor fall on the same rotational face of the helix and can contact, or on opposite faces and cannot. This anchor-relative contact read is present for every locus, and the global periodicity remains valid only as a descriptive composition fact, not as a coordinate claim.

Stress test (v1.9) — contact_competent is at chance. Across the twelve cross-kingdom regions, the rate at which a real TSS sits on the same helical face as its nearest anchor is statistically indistinguishable from chance: pooled over 246 motor–anchor pairs the same-face rate is 0.337 versus an analytic chance of 0.340 (permutation z = −0.09). Real transcription starts are therefore not helically phased to the composition-shell anchors. The flag is read as deterministic Layer-1 geometry, not a functional-contact claim — and the result vindicates the two-layer thesis, since realized looping depends on runtime factors (bound factors, cohesin) that are Layer-2. The genuinely real helical signal is the global nucleosome-positioning propensity (the ~10.5 bp WW periodicity, elevated above shuffle in all twelve), a descriptive composition fact. geometry / chance-level stress test → repro

Methylation kept, and promoted over raw density

Raw CpG density counts CG steps but rises and falls with GC content. The clade reader normalizes it as observed-over-expected, and the difference is large: the same lactase promoter reads a raw density of 0.0148 but a CpG O/E of 0.335, the value that actually reflects depletion. The unified engine demotes raw density to a secondary note and makes O/E the primary environment measure, then routes the locus through the four-regime detector.

The whole clade layer is reproduced exactly — fifty-two quantities across human, six plants, and six insects, with the auto-detector still routing every regime correctly. A negative control confirms the signal is structure, not composition: a shuffle that preserves base counts but destroys dinucleotide order erases the depletion, moving the promoter O/E from 0.335 to 0.998 and dropping the global CG regime.

What stays immutable, and what no gray zone means

The material γ is not touched. It is imported from the locked tables, and a gate re-derives human_SOX2 at 1.287315 every run; the two NN tables behind it are identical. The switch, the coordinate constants, and the methylation thresholds are all locked, so this chapter adds a unifying read without moving any anchor beneath it.

No gray zone means nothing is silent, not that there are no unknowns. The runtime sign, fill, and role are flagged as Layer-2; the sparse-annotation loop is flagged as open and names its closing dataset; the retired helical claim is logged with its reason. The cross-reference belongs in the standing retired register of §8, an author edit left for that chapter, which is locked in this build.