4D DNA Blueprint #7 — The experiment built to break the theory found a size signal in one switch

[ Where we are ]

We have a reading method (Parts I-II) and a bold claim that has run underneath all of it: the genome encodes form, not amount — and body size is amount. A claim is worth little until it is put at risk. This episode designs the experiment that could destroy the theory and runs it on real genomes. The result is more interesting than a clean yes or no: the theory does not break, but it is forced to get sharper, and we have to retract one tempting conclusion along the way. By the end, size is not pure runtime after all — one piece of it turns out to be written in the code, in a single switch, as a direction.

7.1 The prediction, sharpened into a risk

The prediction the cliffhanger promised. The genes that build the size machine — the ones that decide how long a bone grows — are the growth-plate genes: IHH, FGFR3, NPR2, PTHLH, SOX9. If the genome encodes body size anywhere, the most natural place is the “shape” of these very genes: their stiffness signal, the gamma we learned to read in #5. A bigger animal would carry a systematically different shape.

So the theory is forced to predict: if size is amount, not code, then the shape of these genes should carry no signal that tracks how big the animal is. Find a clean size signal in the shape and “quantity is runtime” is in trouble. Set it up that way, and a single measurement can move the theory.

7.2 The first look: the whole-gene shape says nothing

Take the cleanest big contrast available — mouse, cat, tiger, and African elephant, across a span of roughly two hundred thousand fold in body mass — and read the gamma shape of each growth-plate gene in each animal, through the same pipeline, with the data locked as

3 demands. Average the shape over the whole gene, account for which gene is which, and

ask whether the leftover tracks body mass.

It does not. The correlation between whole-gene shape and (log) body size is about +0.003

• which is to say, zero. Measured this way, there is no size component hiding in the shape. This is exactly the result the early versions of this work reported, and it looks like a clean win: size is not in the code.

It was the right measurement of the wrong window.

7.3 We were looking in the wrong place

The theory itself tells us where to look, and it is not at a whole-gene average. The growth switch from earlier in the series — the element that decides whether a program is on or off — does its work at the switch element: the promoter, the short stretch of cis sequence that sets the threshold. A whole-gene average blends that switch in with everything around it and dilutes exactly the signal we are hunting.

So measure gamma where the model says it acts: at the promoter, a small window just around each gene’s start. Across the same four animals, the verdict flips. Promoter gamma tracks the size direction — the larger animal tends to carry the stiffer switch — at a correlation around +0.70. The signal survives a control for local GC content (the promoter’s excess over its own neighbourhood still tracks mass, around +0.78) and it is localized to the promoter (the surrounding hundred-plus kilobases are null, so this is not some genome-wide drift). The whole-gene zero was a fact about the wrong window, not about the biology.

(One honest correction belongs here. The earlier pass had two of these genes — FGFR3 and NPR2 — swapped in the cat and tiger annotations. The gene-feature coordinates give the correct assignment, used from here on. We keep the fix visible rather than quietly absorbing it — that is the rule from #3, not an afterthought.)

7.4 The trap: a tempting law that did not replicate

Now the method’s honesty machinery earns its keep. A correlation of +0.70 across four species looks like a size law. But two of those four — cat and tiger — share a lineage, so this is really closer to three independent points than four. Article IV of our method (no claim rests on counting resemblances) and Article III (a predicted negative is a result; corrections stay visible) both demand a real replication before we believe it.

So we ran one, in a lineage that does not overlap with the first: five primates, from a sixty-gram mouse lemur to a hundred-and-sixty-kilogram gorilla. The correlation did not replicate. It inverted — to about -0.47. The diagnosis is clean, and it teaches the real lesson. Each lineage sits at its own baseline stiffness, and those baselines are not ordered by size: a lemur’s switches sit about as stiff as a tiger’s and an elephant’s, with the mouse lowest of all. In the original four animals the lineage baselines happened to line up with body size; primates break the coincidence. So the cross-lineage +0.70 was largely a between-lineage baseline alignment, not a size code — and we withdraw it as a body-size law. (It now lives in the paper’s failure log: not buried, relocated, so the correction stays on the record.)

This is the part of the story the old version of this episode never told, because the old version stopped at the clean zero. The honest path is messier and better: a tempting result, put at real risk, failed — and that failure pointed straight at the actual signal.

7.5 What survives is narrower, and much stronger

Strip out the lineage baseline and two independent within-lineage contrasts remain: cats (house cat to tiger, a fifty-five-fold jump) and great apes (human to gorilla, a 2.6-fold jump). In both, the larger animal’s growth switches are stiffer. And when you decompose that shift gene by gene, almost all of it sits on one gene — the same gene in both lineages.

That gene is PTHLH. It is the growth plate’s duration timer: it keeps the pool of dividing cartilage cells going and delays the moment they stop — which is to say it sets how long the plate stays open and how many division cycles accumulate. In the cats and in the apes

• two completely independent evolutionary routes — the smaller species sits at the same switch stiffness and the larger sits higher by the same amount: a shift of about +0.135 in gamma, in the same gene, in the same direction. Pool all nine animals across lineages and PTHLH shows nothing (correlation about +0.006). Its signal is purely within-lineage, which is the fingerprint of a real growth variable rather than a between-lineage artifact.

We did not stop at correlation. Feeding the measured switch stiffness into the development engine and intervening directly on PTHLH — clamp its contribution and the size ordering of the species collapses; restore it and the ordering returns — shows the gene is carrying the direction causally, not just correlationally. One switch, read as a dwell time, sets which animal is larger.

7.6 Direction, yes. Magnitude, still open — and not handed to “runtime”

Two honesties keep this from becoming an over-claim.

First, this is a direction code, not a magnitude code. A fifty-five-fold size jump (cats) and a 2.6-fold jump (apes) produce the same +0.135 shift. The switch says which animal is bigger; it does not say by how much. The whole spread of switch stiffness across species is only about six percent against a two-hundred-thousand-fold range of mass, and inside the cat family the difference is essentially flat. So gamma owns the relative direction of how long the plate stays open — a genuine quantitative share of size, through the duration factor — but not the absolute scale.

Second, and this is the real change from the old version: the magnitude is not handed to “the runtime.” Saying “size is runtime” names no mechanism; it is a place to hide an unknown. Body size is, near enough, accumulated cell number, and cell number is (proliferation rate) times (how long growth stays open). The duration factor is the switch dwell we just measured — that part is in the code. The rate, and the absolute calibration of both, are set by named genomic mechanisms that are real and citable: the let-7/HMGA2/LIN28B developmental timing clock (whose growth brake we will count in a later episode — it is conserved across a thousand-fold range of size, so the size difference lives in its timing, not its presence); the GH/IGF1 dosage axis (a single IGF1 variant is the main reason some dog breeds are small); and a high-dimensional polygenic background (human height alone draws on roughly twelve thousand common variants, concentrated in growth-plate genes). All of these are in the genome. None of them is “runtime” used as a shrug.

7.7 Why this is the method working, not a retreat

It would have been easier to leave this episode as it first stood: a clean zero, theory confirmed, move on. The honest version is better. The original instinct survives — size is not a simple number stamped on a gene, and the whole-gene zero reproduces exactly. But by measuring where the model said to look, by killing a false law the moment it failed to replicate, and by decomposing down to the one switch that actually carries the signal, the theory came out sharper than it went in. A reading tool that never has to revise itself is not a tool; it is astrology. This episode is the machinery that keeps it a tool.

7.8 Try it yourself (a design exercise)

You probably cannot pull elephant and gorilla promoters on a whim, so do the design instead. Pick a trait you think might be “in the genome” — size, lifespan, speed. Now ask two questions, not one. What exact quantity, read off the sequence, would have to differ between a large and a small species? And — the lesson of this episode — at which window would you measure it? A whole-region average can hide a signal that lives in a short switch; the wrong window returns a confident “absent” for a thing that is really there. Write the quantity and the window down precisely enough that a measurement could come back either way, and you have turned a belief into a test.

— WHERE THIS STOPS — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — This is a direction code, recovered within two independent lineages (cats and great apes) and localized to a single switch, PTHLH. It is not a magnitude law, not a universal size code, and not a claim that the genome stores how big an animal will be. The cross-species correlation that first looked like a size law is withdrawn: it was a between-lineage baseline coincidence, and an independent order of animals inverts it. The support is two lineages; the decisive next test is a single close-related group with four or more species. And the magnitude itself — by how much — is still open: it is attributed to named genomic mechanisms (a timing clock, hormone dosage, a polygenic background), but decoding those quantitatively is work not done here. — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —

— CHECK IT IN THE PAPER — — — — — — — — — — — — — — — — — — — — — — — — — — — — — - This episode is the teaching version of the whitepaper’s switch-element size result and its correction: the whole-region gamma null is reproduced, the size direction is recovered at the promoter, the cross-lineage correlation is withdrawn as baseline-confounded, the surviving within-lineage signal localizes to PTHLH, and the magnitude is attributed to named genomic mechanisms rather than the runtime. The relevant modules are biggest_animals, growth_causality_v10, and form_switch_4d_v10. Full paper, proofs & reproducibility bundle (always-latest, now v11): https://doi.org/10.5281/zenodo.20471407 — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — —

Next up — #8: The missing middle. The genome does not touch the body directly — it is read into proteins, and proteins do the work. We put that middle layer under the same reading, across animals from a mouse to a whale, and ask what the proteins keep the same and what they leave for the cell to decide.

— — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — — Part of the 4D DNA Blueprint series. Full paper, proofs & reproducibility bundle (always-latest): https://doi.org/10.5281/zenodo.20471407 Project: https://jamming-physics.org/ 4D DNA Blueprint — what the sequence fixes, and what it does not. © 2026 Young Jae Lee — CC BY 4.0