The grammar hierarchy: reading the upper blueprint

Everything the corpus has read is the cell level: the material γ and the A4 coordinate come from the base sequence — the chemistry of adjacent base pairs, the bottom blueprint. But the heart forced us to import external measured moduli, because its arrangement was orthogonal to that cell-level material. So is there really no γ or A4 at the tissue level, or have we just not yet seen it? This appendix finds the higher blueprints. The same level-and-shape operator that gives the cell its γ and A4 is lifted up a tower of three grammatical levels: the material grammar (cell, the physical substance of the letters), the regulatory grammar (tissue, the arrangement of transcription-factor binding sites — the cardiac code GATA, NKX2-5, MEF2, TBX5, HAND), and the architecture grammar (organ, the layout of those elements — the arrangement map). On real human promoters the regulatory grammar reads cardiac tissue identity from sequence — cardiac promoters carry roughly fourteen times the cardiac binding-site grammar of a housekeeping control — while the material grammar is blind to that identity. A dinucleotide-preserving shuffle preserves the material level exactly yet destroys three-quarters of the cardiac grammar, proving the tissue identity is information above the material level entirely. The corpus had been reading the wrong level of the blueprint; the arrangement the heart needed was written one grammar up. This is not the whole blueprint solved — the functional validation and the real arrangement map stay open — but the upper γ and A4 are found, formalized, and read from sequence.

The cell-level reading treats one signal — the nearest-neighbour base-stacking free energy, the electrostatic-chemical material of adjacent letters — and splits it into a LEVEL, the material γ that says how stiff the cell's material is, and a SHAPE, the A4 coordinate that says where along the sequence the material varies. That is the bottom blueprint, and it is all the corpus has read; when Appendix A asked it to set the heart's developmental arrangement it returned an orthogonal null, and Appendix D had to import measured cardiac moduli to proceed. The user's instruction names the gap exactly: if the blueprint is precise, the heart's exact form cannot come from cells alone — there must be a tissue-to-tissue arrangement map — and if the A4's material began as ions and chemistry, the upper blueprint's material must be something else. It is. This appendix formalizes a hierarchy of grammars, each reading a higher-order signal with the identical level-and-shape operator, so that the cell's (γ, A4) is the bottom rung of a tower. The material grammar G1 reads stacking energy, phonology, the substance of the letters, and returns the cell's (γ₁, A4₁). The regulatory grammar G2 reads the cis-regulatory syntax, the arrangement of transcription-factor binding sites — for the heart the established cardiac code of GATA4, NKX2-5, MEF2C, TBX5 and the bHLH HAND factors — and returns a tissue-level (γ₂, A4₂) whose shape is the arrangement of cardiac control along the sequence, the upper A4 the corpus had not formalized. The architecture grammar G3 reads the layout of those regulatory hotspots, discourse, how the whole text is organized, and returns an organ-level (γ₃, A4₃), the arrangement map. Three results, on real human promoters embedded for offline reproduction — two cardiac, NPPA and the cardiac troponin T promoter, and one housekeeping control, GAPDH — establish that the upper grammars are real and were missed. First, the regulatory grammar separates cardiac from housekeeping by about fourteenfold while the material grammar does not carry that identity, so a tissue-level blueprint exists and is readable from sequence. Second, a dinucleotide-preserving shuffle, which preserves the dimer counts and therefore the material level γ₁ exactly, collapses three-quarters of the cardiac grammar, so none of the tissue identity lives at the material level. Third, the material and regulatory signals are essentially uncorrelated along the sequence, distinct projections of the blueprint rather than the same signal twice; and the shape projection is the identical robust operator at every level, the cell-level grammar lifted, not a new rule per rung. This locates, in the sequence, the arrangement the heart forced Appendix D to import: tissue identity and the arrangement signal become sequence reads, reducing the external-data reliance, while the absolute kilopascal magnitudes and the functional phenotype remain measured. Every constant is locked with a grade and provenance, zero inline magic numbers; the fail-closed gate passes nine of nine deterministically; and completion is honestly false — the upper blueprint is found and formalized, its functional and three-dimensional-arrangement validation named as the next obstacle.

Why this appendix exists

The corpus reads DNA at one level. The material γ and the A4 coordinate are both computed from the base sequence — γ is the mean nearest-neighbour stacking free energy, the chemistry of how adjacent letters stack, and A4 is the pattern of that same material with its mean removed. That is the cell's blueprint, the bottom of the structure, and it is genuinely all that has been read. Its limit showed at the heart: Appendix A's gene-clock found the material orthogonal to the heart's developmental arrangement, a null sharpest in the heart, and Appendix D could only proceed by importing measured cardiac moduli. The user's instruction names the missing piece precisely. If the blueprint is precise rather than accidental, the heart's exact form cannot emerge from cells alone — there must be a map of how tissue is arranged against tissue — and if the A4's material began as ions and chemistry at the bottom, then the upper blueprint's material is not that, but something higher. This appendix asks what that something is, and whether it is already written in the sequence at a level the cell-level reading skips.

The hierarchy of grammars

The answer is a hierarchy. The cell's level-and-shape operation is not specific to stacking energy; it is a way of reading any signal as a level and a shape. So it can be lifted. At the bottom, the material grammar reads the stacking signal and returns the cell's γ and A4 — in linguistic terms this is phonology, the physical substance of the letters, the sound and stuff of them. One level up, the regulatory grammar reads not the chemistry of adjacent bases but the arrangement of functional words: the binding sites of transcription factors, whose spacing, orientation and combination form what genomics already calls the cis-regulatory grammar. For the heart that grammar is the established cardiac code — GATA4, the NK homeodomain factor NKX2-5, the MADS-box factor MEF2C, the T-box factor TBX5, and the basic helix-loop-helix HAND factors — whose combinatorial syntax specifies cardiac enhancers. This is syntax, how words combine, and its level-and-shape split returns a tissue-level γ that measures the overall cardiac regulatory drive and a tissue-level A4 whose shape is the arrangement of cardiac control along the sequence — the upper A4 the corpus had not formalized. One level higher again, the architecture grammar reads the layout of those regulatory hotspots, how they are spaced and ordered, the discourse structure of the whole text; its A4 is the arrangement map, the layout that becomes the spatial arrangement of tissue. The canonical sequence-encoded arrangement map is Hox colinearity, where linear order on the chromosome is spatial order along the body; here it is read locally as the spacing of the regulatory elements.

The upper material is syntax, not chemistry

The user's sharpest question is what the upper blueprint's material is, if at the bottom it was ions and chemistry. The hierarchy answers it directly. The material of the bottom grammar is physical: the stacking free energy, electrostatics and base-stacking, the chemistry of the letters. The material of the regulatory grammar is not physical in that sense — it is informational, the syntax of control, which transcription factors are written where and how their sites are arranged. The material of the architecture grammar is higher still, the layout itself, the spatial organization of the regulatory text. Each level has its own material and its own level-and-shape reading, and the materials are not the same kind of thing — chemistry, then syntax, then architecture — which is exactly why reading only the chemical material at the bottom cannot recover the tissue arrangement written in the syntax above it.

Tissue identity is readable from sequence

The first result is that the tissue-level blueprint is real and readable. On real human promoters — two cardiac, the natriuretic peptide gene NPPA and the cardiac troponin T gene, and one housekeeping control, GAPDH, all unmodified GRCh38 sequence — the regulatory grammar separates them cleanly. The cardiac promoters carry on the order of eleven to seventeen cardiac transcription-factor binding sites, the housekeeping promoter just one, a separation of about fourteenfold. The material grammar, by contrast, does not carry that identity: the material γ is nearly the same across all three, within a few percent, because the chemistry of adjacent bases is similar whether or not the region is cardiac. So which tissue a region builds is written in the sequence, but at the regulatory level, not the material level — and the corpus had been reading the material level. The contrast is not circular, because these are real genomic sequences with their real binding-site content, not constructed examples.

The tissue identity is information above the material level

The second result makes the separation exact. A dinucleotide-preserving shuffle rearranges a sequence while keeping the count of every adjacent-letter pair identical. Because the material γ is nothing but the mean of those pairwise stacking energies, the shuffle preserves the material level exactly — to the last digit, not approximately. Yet the same shuffle collapses the cardiac grammar, destroying about three-quarters of the cardiac binding sites, because the specific arrangements of letters that form the binding words are broken apart even though the letter-pair counts are unchanged. The conclusion is sharp: none of the tissue identity lives at the material level. Everything that distinguishes a cardiac promoter from a scrambled sequence with identical material is in the higher-order arrangement. Reading only the cell-level material was guaranteed to miss it, and the shuffle proves the missing is total, not partial.

The levels are orthogonal, and the operator is one

The third result confirms the levels are genuinely distinct. Along the real sequence the material signal and the regulatory signal are essentially uncorrelated, a correlation near minus four hundredths and a coefficient of determination near two thousandths; they are different projections of the same blueprint, not the same signal read twice. And the operation that produces the shape at each level is the identical robust operator the cell uses for the A4 coordinate — median-centred, scaled by the median absolute deviation — invariant to rescaling and to shifting the signal to within machine epsilon. So the three grammars are not three ad hoc rules but one grammar applied at three levels, the cell's reading lifted up a tower from phonology to syntax to discourse.

The heart resolution, and what stays measured

This locates, in the sequence, the arrangement the heart forced Appendix D to import. Appendix D bracketed the myocardium with measured cardiac moduli because it read the cell-level material, which is orthogonal to the arrangement; the arrangement lives one grammar level up, in the cardiac transcription-factor syntax, and is sequence-readable. The blueprint had the arrangement all along — the corpus had been reading the wrong level. Each grammar supplies a different part of the heart: the material grammar supplies the cell's intrinsic stiffness, read before but blind to identity; the regulatory grammar supplies which tissue, the cardiac identity; the architecture grammar supplies the spatial layout, toward the organ's exact form that cells alone cannot give. The external-data reliance is reduced accordingly: tissue identity and the arrangement signal become sequence reads rather than imports. What remains measured is honest and named — the absolute kilopascal magnitudes, which the regulatory grammar does not set and which stay the measured inputs of Appendix D, and the functional phenotype, the validation that the arrangement predicts measured enhancer activity, which is the next obstacle.

What is locked, and what stays open

Every input — the stacking parameters, the cardiac transcription-factor consensus motifs, the three real promoter sequences with their accessions and coordinates, the shuffle seed — is read from the locked database with a grade and a provenance, and the lock manifest reports zero inline magic numbers. The decompositions and invariances are verified-exact: the same operator at every level, the exact preservation of the material level under the dinucleotide shuffle, the orthogonality of the levels, all reproduced deterministically under a double SHA-256. What stays open is the accuracy, and three channels carry named obstacles. The functional validation, that the regulatory shape predicts measured cardiac enhancer activity, needs a held-out functional dataset — the cardiac enhancers of the VISTA database, cardiac open-chromatin or active-histone marks, or a massively parallel reporter assay — tested with a pre-registered rank statistic and a shuffle control. The real arrangement map, the true spatial layout of tissue, needs measured three-dimensional genome contacts or the colinear spatial-domain transcription-factor map, staged across development. And the absolute moduli remain the measured inputs of Appendix D, which this appendix does not supersede; it reduces the identity and arrangement import, not the magnitude import. Until the functional dataset is supplied and the pre-registered test is run, completion is false. The upper blueprint is found and formalized — its quantitative validation is the next step.