Autism as a three-axis fault — threshold, gain, and wiring

Autism is modeled not as one lesion but as three separable faults on the emerged cerebrum: a threshold E-I fault, an output gain fault, and a long-range wiring fault. Each leaves a distinct, reproducible fingerprint in synchrony, theta-gamma coupling and ignition; which fault an individual's autism is stays open. efficacy=0; not medical advice.

This chapter applies the paper’s pathology faculty (F10) to autism on the same emerged, measured cerebrum the rest of the paper builds. Autism is read not as one lesion but as three separable faults on the coordination substrate: a threshold (E-I) fault T, an output/gain fault O, and a long-range wiring fault W. Seventeen real autism risk genes, each carrying a measured stacking-energy γ, sort onto the three axes, and each axis leaves a distinct, bit-reproducible fingerprint in global synchrony, θ–γ coupling, and ignition. The model asserts the fingerprints; which axis any individual’s autism is remains open.

Why three axes and not one

Autism is behaviourally one diagnosis but mechanistically heterogeneous, so the model does not posit a single broken part. It takes the emerged cerebrum of §13 — twelve central organs coupled only by the measured ephaptic near-field at the measured strength κ = 0.5496 — and asks the narrow question it can answer: in how many mechanically distinct ways can that coordination fail? The answer the dynamics give is three, and they are separable because they act on different terms of the same coupling.

A threshold (T) fault raises the excitability fold — the E-I balance shifts so an assembly needs more drive to ignite. An output (O) fault lowers the drive each node delivers (the coupling numerator falls) without moving the fold. A wiring (W) fault leaves both the fold and the per-node drive intact but changes the geometry: local edges over-connect and long-range edges under-connect. The three are not three names for low synchrony; they are three different operators on the same equation, and that is why a single readout cannot tell them apart but two can.

The cohort: seventeen risk genes, each with a measured γ

The axes are not assigned by hand. A pre-registered cohort of seventeen autism risk genes — fifteen drawn from the verified risk-gene atlas and two fetched live from NCBI RefSeq (GABRA5, MACROD2) — is read through the same pipeline the organs use: the promoter window [TSS−2000, TSS+500] (2501 bp, coding strand) gives a mean nearest-neighbour stacking energy γ by SantaLucia 1998. The cohort is deliberately moderate-or-below: severe developmental and epileptic-encephalopathy syndromes (CHD8, FOXG1, FMR1, GRIN2B, DYRK1A and the like) are pre-registered out, because forcing a single profound-pole master onto the cohort would overclaim. What is kept is the legible, axis-mappable end.

Each gene’s mechanism places it on one axis: the T axis carries the ion and GABA-A levers (CACNA1C, GABRA2, GABRA5, GABRB3, KCNQ3, SCN2A); the O axis carries the synaptic-gain scaffolds and splice/output regulators (CTTNBP2, MACROD2, NRXN1, RBFOX1, SHANK2, SHANK3, SYNGAP1); the W axis carries the adhesion and lamination genes that build long-range connections (CNTN6, CNTNAP2, DSCAM, RELN). Mapping mechanism to axis is the only interpretive step; the γ values themselves are measured and re-derive offline bit-for-bit.

The fingerprint: what each fault does to the field

Run on the cohort cerebrum, the three faults all pull global synchrony below the healthy metastable value (R_health = 0.390): the output fault settles at R = 0.354, the threshold fault at R = 0.366, the wiring fault at R = 0.340. If R were the only instrument, the three would be a smear. They are not, because θ–γ phase–amplitude coupling (PAC) and the ignition threshold separate them.

The decisive observable is the pair (ΔPAC, ignition). The O fault drops PAC and keeps the ignition fold normal. The T fault drops PAC and raises the ignition fold — the signature of a shifted E-I balance. The W fault is the clean one: its PAC is unchanged, equal to health to within the simulation’s precision, and its fold stays normal, while a locality imbalance (local-over, long-range-under) appears that the other two do not have. So ΔPAC < 0 with a raised fold is T; ΔPAC < 0 with a normal fold is O; ΔPAC = 0 is W. The fingerprint is exact and reproduces under SEED = 19.

What is forced, and what stays open

Two things are owed in the open register, and naming them is the same no-tuning discipline that lets the rest stand. First, O and T are degenerate at the coupling level — both scale the same κ — and separate only at the fold; distinguishing them in a real brain therefore needs an excitability or ignition measure (a TMS–EEG threshold), not a spectral one alone. Second, and more important, the model asserts the fingerprints and their reversibility, not which fault any individual’s autism is. That mapping requires per-person external data — connectome, spectrum, genetics — and is held [O].

This three-axis picture is the foundation the next chapters stand on. It already implies the clinical asymmetry the rest of Part II explores: because O and T are gain faults and W is a geometry fault, a chemical that shifts gain will behave very differently on the three — reaching some and only masking others (§19). Every fraction and coupling value here is an in-silico coupling state, not a clinical response rate or a dose; efficacy = 0 throughout. This is a mechanism-level result about the fault structure of autism as represented in the VP framework — not medical advice, a diagnosis, a treatment protocol, or a cure.