Autism threshold levers — the E/I over-excitation operating-point correction, decomposed into DNA-grounded levers (L1-dominant; the output and wiring axes named, out of reach)

Chapters 18-19's E/I over-excitation correction is decomposed into a DNA-grounded three-lever target map: reduce inward excitatory current (L1, dominant), increase outward K+/restore GABA-A inhibition (L2), with a sparse serotonergic lever (L3), over ten autism genes. The map reaches the excitability axis only; the named output and wiring axes stay out of reach; targets ranked, never drugs; not medical advice.

The autism chapters — the three-axis chapter and the chemical-limits chapter — read autism not as one severity dial but as three distinct fault axes: T, an excitability / E-I threshold axis (an over-excitation operating point whose ignition fold sits too low); O, a synaptic output / gain deficit; and W, a long-range wiring / connectivity axis. They proved the pole and the corrective sign for the reachable axis — the T axis is an over-excitation pole, so the corrective push is whatever reduces the excess drive or restores inhibition (raises the fold), the same direction as epilepsy and the schizophrenia positive domain — but they carried a second, sharper result the symptom-checklist view cannot: a scalar chemical lever reaches the T axis only, and §19 proved the W axis is unreachable — a threshold shift can only mask a wiring fault by over-synchronisation (the seizure analogue), never correct it. This chapter does not invent a new map; it unifies the pre-existing autism multilever module under the same piece of inherited technology the bipolar, epilepsy, depression and schizophrenia levers chapters used: the threshold-shift intervention logic from the analgesic reproducibility package (Zenodo 10.5281/zenodo.20733420). Its three abstract levers — L1 change the inward (excitatory) current, L2 change the outward (potassium / inhibitory) current, L3 change the up-stream drive — carry over unchanged, and autism is the fifth distribution pattern across the series: it is L1-dominant with a nearly-empty L3. Of ten lever genes, five sit on the excitatory-reduce lever (L1: GRIN2A, GRIN2B, GRIA1, SCN2A, CACNA1C), four on the inhibitory-restore lever (L2: KCNQ3, GABRB3, GABRA5, GABRA2), and just one on the up-stream lever (L3: SLC6A4, a cautious, non-monotone serotonergic node graded [O]) — the L3-sparsity is itself the finding, that autism's actionable biology sits locally on the excitation/inhibition set with no clean up-stream drug drive. Two structural strengthenings set this chapter apart from the schizophrenia domain-restriction template. First, the out-of-reach axes are named with real genes: the O (output-deficit) axis carries SHANK3, SYNGAP1, NRXN1; the W (long-range-wiring) axis carries CNTNAP2, RELN; and the syndromic master MECP2 rides alongside — carried on the map but explicitly not levers. Second, the W-axis unreachability is proven, not asserted: §19 showed a scalar threshold lever can only mask the wiring fault, never correct it. Each gene is placed by reading its own promoter switch stiffness — γ = −mean nearest-neighbour stacking free energy (SantaLucia 1998) over its promoter window, turned into |h_sp| = spinodal(γ) with the frozen engine read-only, and seven of the reads (GRIN2A, GRIN2B, CACNA1C, GABRB3 from the schizophrenia cache, SCN2A, KCNQ3 from the bipolar cache, SLC6A4 from the depression cache) are carried over verbatim. Two honest caveats are recorded, not hidden: SCN2A and GRIN2B are gain/loss sign-subtle (a gain-of-function pushes the early-infantile developmental-epileptic-encephalopathy / seizure pole while a loss-of-function pushes the milder autism / intellectual-disability pole — opposite directions), and the threshold lever pushed too hard is itself the seizure edge (the real autism+epilepsy comorbidity). Three fail-closed disciplines ride along: every channel and serotonergic link stays graded [O] cited biology, a forbidden-claim scanner rejects any dose / efficacy / safety / synthesis statement and — uniquely for this topic — an autism quackery class (chelation, bleach / miracle-mineral protocols) and a normalise-framing class that protects neurodiversity respect, and a burden-weighted ranking orders targets, never drugs or doses. The firewall is absolute: the promoter |h_sp| is a gene's own switch stiffness, never the §18 network over-excitation threshold, a receptor occupancy, a synaptic level, a potency, a dose, or a clinical effect. efficacy = 0; not medical advice; autism is a neurodevelopmental difference, not only a deficit; the hard problem stays open.

What §18-19 left abstract (and what they proved cannot be reached)

The two autism chapters — the three-axis chapter and the chemical-limits chapter — did something the symptom-checklist view of autism cannot: they replaced a single severity dial with three distinct fault axes, each a different kind of difference. The T axis is excitability — the excitation/inhibition (E-I) balance — and on the atlas it is an over-excitation operating point: health is a population of micro-eddies igniting in gamma only when activity clears a threshold (the ignition fold), and the T-fault is that fold sitting too low, so circuits over-respond. The O axis is an output / gain deficit at the synapse — too little effective transmission, a different failure entirely. The W axis is long-range wiring — a connectivity / geometry problem, local circuits over-connected and long-range links under-connected. These are not three severities of one thing; they are three axes that can vary independently, which is why autism presents so differently from person to person.

§18-19 proved two things about these axes and stopped. The first is the pole and sign of the reachable one: the T axis is an over-excitation pole — the same side of the line as the over-synchronisation disorders — so the corrective sign is whatever reduces the excess drive or restores inhibition (raises the fold), the mirror of depression's restore-the-deficit and the twin of epilepsy's and the schizophrenia positive domain's reduce-the-excess. But §18-19 left that push abstract. The second result is the reason this chapter is unlike the four before it: §18-19 also proved a reach limit. A scalar chemical lever — a drug that raises or lowers a current or a drive — can act on the T axis, but it cannot lift the O deficit (a gain-reducing lever pushes a deficit further down), and §19 went further and proved the W axis is unreachable: a threshold shift can only mask a wiring fault by forcing over-synchronisation — the seizure analogue — and can never re-route the geometry that the fault lives in. A mechanism atlas should be able to say which molecular targets realise the T-axis correction and be honest, by name, about the axes a chemical lever cannot reach. That is exactly what this chapter does.

The inherited technology, applied a fifth time — as a unification, and the L1-dominant pattern

The handle is not invented here and it is not adapted here — and in autism's case it is not even newly applied. Autism already had a multilever threshold module that carried this exact idea under informal A1/A2/A3 labels. What this chapter does is unify that module under the same formal frame the bipolar, epilepsy, depression and schizophrenia chapters inherited from the analgesic reproducibility package (Zenodo 10.5281/zenodo.20733420). Its premise is general: an operating point is set by a balance of currents and the drives that bias them, so there are exactly three levers on it. L1 — change the inward, excitatory current. L2 — change the outward, repolarising (or inhibitory) current. L3 — change the up-stream drive that sets where the operating point sits. The frame applies unchanged because the autism chapters, the four levers chapters, and the analgesic package all share the same R19 substrate — the engine's supercritical pitchfork ṣ = g·s − s³ + h, whose spinodal fold IS the switching barrier — so the informal A1/A2/A3 levers and the formal L1/L2/L3 levers are the same three operators, and the unification loses nothing.

Autism is the fifth distribution pattern across the series, and a distinctive one. Where bipolar leaned on L1 (calcium channels), epilepsy on L1+L2 (the KCNQ2/KCNQ3 M-current), depression on L3 (the up-stream HPA / monoamine / neurotrophic drives, L3-dominant), and schizophrenia on L1+L3 co-dominantly, autism is L1-dominant with a nearly-empty L3: of the ten lever genes, five sit on L1, four on L2, and just one on L3. That single-gene L3 is not an omission — it is a finding. Autism's replicated genetic signal is overwhelmingly in the excitation/inhibition machinery itself (glutamate and GABA receptors, voltage-gated channels), not in a clean up-stream neuromodulatory drive the way depression's HPA axis is; so the map reports, faithfully, that autism's actionable biology sits locally on the E/I set, with no clean up-stream drug handle. The pipeline is reused at the level of code, not analogy: seven of the reads — GRIN2A, GRIN2B, CACNA1C, GABRB3 from the schizophrenia cache and SCN2A, KCNQ3 from the bipolar cache and SLC6A4 from the depression cache — are carried over verbatim, because γ is a strand-symmetric property of the sequence and does not change between problems. Nothing about the engine is touched; the module re-emerges the frozen tree read-only and confirms it byte-unchanged, and registers as the twelfth atlas citizen (ASD-T-L).

The new finding: a T-axis map whose out-of-reach axes are named (and one is provably unreachable)

This is the result that makes autism the most structurally explicit chapter in the levers series. The schizophrenia chapter introduced domain restriction — the lever map reached the positive domain only — but it left the unreached domains as categories. Autism does two things schizophrenia did not. First, it names the out-of-reach axes with real genes, and carries them on the map as explicitly non-levers. The O axis (synaptic output / gain deficit) carries SHANK3 (the postsynaptic master scaffold of Phelan–McDermid syndrome), SYNGAP1 (a synaptic Ras-GAP gain regulator), and NRXN1 (presynaptic neurexin-1). The W axis (long-range wiring) carries CNTNAP2 (Caspr2, a long-range cell-adhesion molecule) and RELN (reelin, the cortical-lamination / migration signal). The syndromic master MECP2 (X-linked, the chromatin / transcriptional regulator behind Rett and MECP2-duplication syndromes) rides alongside as a whole-program perturbation. None of these is a lever, because a gain-reducing scalar push would drive an O-axis output deficit lower, not higher, and cannot re-route a W-axis geometry at all.

Second, and more sharply, the W-axis unreachability is not a hedge — it is a result the model imports as proven. §19 showed, on this same substrate, that a scalar threshold lever applied to a wiring fault can only mask it by forcing the network into over-synchronisation — the very seizure analogue the T lever pushed too far produces — and can never correct the geometry (P4_chemical_cannot_fix_W, P5_threshold_lowering_is_mask_not_correction). So the lever map reaches the T axis and only the T axis, the O and W axes are named but unreached, and the W axis is provably unreachable. The module records this in a domain-restriction witness (T = reached, O = not reached, W = not reached) and a separate out-of-reach-targets section that lists the six genes by axis. Capturing the partiality precisely — axis-structured and gene-named, not dose-structured — is far more informative than a map that pretended to cover all three, and it is why the fail-closed L3-honesty gate for this chapter checks not only that the serotonergic link stays [O] but that L1 is the unique dominant lever, that L3 is sparse, that the T-axis restriction holds, and that at least one O gene and one W gene are named and flagged unreached with the §19 citation; it FAILS the build if any of these slips.

L1 (dominant) — the inward excitatory-reduce levers (and the gain/loss seizure-edge subtlety)

The dominant lever is the excitatory-reduce axis, and it carries the chapter's most important honest caveat. Five genes sit here. GRIN2A and GRIN2B are the NMDA-receptor GluN2A and GluN2B subunits (glutamate, calcium-permeable); GRIA1 is an AMPA-receptor subunit carrying the fast-excitatory sub-route; SCN2A is the voltage-gated sodium channel Na_V1.2, one of the most recurrently mutated autism genes; and CACNA1C is the L-type calcium channel Ca_V1.2, the most-replicated cross-disorder locus (the same one that anchored the bipolar and schizophrenia L1 axes). The naive reading of L1 would be reduce inward excitatory current — raise the fold, stop the over-excitation — but in autism the sign is not that simple for two of these genes, and the model flags it explicitly. SCN2A and GRIN2B are gain/loss sign-subtle: a gain-of-function variant pushes the early-infantile developmental-epileptic-encephalopathy / seizure pole, while a loss-of-function variant pushes the milder autism / intellectual-disability pole — the same gene produces opposite clinical directions depending on the variant, so ‘reduce excitation’ is not a clean direction for it. This is the autism analogue of depression's ketamine caveat and the schizophrenia NMDA-hypofunction caveat, and it has a second, autism-specific face: the T lever pushed too hard is itself the seizure edge. Drive the over-excitation correction too far and the network tips into the over-synchronisation pole — exactly the autism+epilepsy comorbidity that is clinically real. So the lever placement (the gene is on the inward-current lever) is structural and trait-blind, but the direction and magnitude of any clinically relevant agent are variant-, dose-, and mechanism-specific and stay graded [O]. A promoter read places a gene on a lever; it says nothing about whether reducing or raising a current is the therapeutic direction, or how far is too far. A direction, never a dose.

L2 — the inhibitory-restore levers (potassium and GABA-A)

The second lever is the inhibitory-restore axis, and it is the restoring mirror of L1 — instead of reducing the inward excitatory current, it raises the outward / inhibitory current that holds the fold up. Four genes sit here. KCNQ3 is the K_V7.3 potassium channel (the M-current), and restoring its outward current is the most direction-consistent move on the whole map: more outward current raises the fold directly, and the channel is druggable (the same M-current axis epilepsy leaned on). GABRB3 (the GABA-A β3 subunit, on 15q11–13, the Dup15q / Angelman overlap that raises its burden), GABRA5 (the extrasynaptic α5 subunit carrying tonic inhibition, with α5-selective ligands already in the literature), and GABRA2 (the synaptic α2 subunit carrying phasic inhibition) raise the inhibitory chloride conductance — the same restoring side of the E/I axis, approached through inhibition rather than through reducing excitation. This is the lever the burden ranking will surface as the cleanest actionable direction, precisely because it does not carry the gain/loss sign problem the L1 channels do: restoring inhibition raises the fold regardless of variant direction. Every claim on this lever is a mechanism direction only; no opener, modulator, dose, or patient is named.

L3 (sparse) — the single up-stream serotonergic node, and why sparsity is the finding

The third lever does not touch a channel; it changes an up-stream drive — and in autism it is almost empty. A single gene sits here: SLC6A4, the serotonin transporter. It is carried as a cautious L3 node and graded [O] for a specific reason: the serotonergic evidence in autism is non-monotone and mixed — it is an adjunct signal, not a clean up-stream correction the way depression's HPA / monoamine drives are. The contrast with depression is the whole point. Depression was L3-dominant: twelve of its eighteen genes sat on up-stream HPA, monoamine, and neurotrophic drives, because depression is a disorder of an up-stream drive biasing an operating point. Autism is the opposite extreme — L3-sparse — because autism's replicated signal sits in the excitation/inhibition machinery itself, not in a clean neuromodulatory drive. The sparsity is therefore not a gap in the search; it is a structural fact about where autism's actionable biology lives: locally, on the E/I set, where the L1 and L2 levers act, with no clean up-stream drug handle to pull. The fail-closed L3-honesty gate enforces both halves of this — that the one serotonergic link stays [O] cited biology (never derived from the substrate), and that L3 is genuinely sparse while L1 is the unique dominant lever. A direction, never a dose — efficacy = 0.

The DNA grounding: a promoter's own switch stiffness, and a decoupling that names itself

What places each of the sixteen genes — the ten levers and the six out-of-reach targets — is not a list but a read. For every gene, the module takes its promoter window (transcription start −2000 to +500 bases, Homo sapiens) and computes γ = −mean of the nearest-neighbour base-stacking free energies along that window (the SantaLucia 1998 nearest-neighbour thermodynamics), then turns that γ into the promoter's switch stiffness through the frozen engine's own functions: |h_sp| = spinodal(γ) = 2(γ/3)^1.5 and barrier = γ²/4. The reads span a real range, and that range tells its own story. The stiffest promoters in the whole set are the out-of-reach genes and the sparse L3 node: the O-axis scaffold SHANK3 at γ ≈ 1.52 (|h_sp| ≈ 0.72), the W-axis lamination gene RELN (γ ≈ 1.51), and the serotonin transporter SLC6A4 (γ ≈ 1.52, |h_sp| ≈ 0.72). The softest reads are the canonical actionable E/I levers: the sodium channel SCN2A at γ ≈ 1.20 (|h_sp| ≈ 0.50, the softest in the set) and the calcium channel CACNA1C (γ ≈ 1.26, |h_sp| ≈ 0.55). In other words, promoter stiffness runs opposite to reachability and actionability here: the genes a chemical lever cannot touch read stiffest, and the genes it can touch read softest. These are read on the same R19 substrate, with the same engine, that the bipolar, epilepsy, depression, schizophrenia and analgesic packages used, which is the whole point of the inheritance: one substrate, one pipeline, now six problems. The γ read is a property of the gene's promoter sequence, blind to whether the gene is on or off and to gain / loss / expression level, and that is all it is.

Ranking targets, the autism unmet-need signature, and the firewall that keeps gamma honest

The last component prioritises, and it prioritises targets, never drugs or doses, over the ten lever genes only — the six out-of-reach genes are not ranked, because ranking a target you have already declared unreachable would be incoherent. A burden-weighted score combines three declared, cited weights — clinical burden (0.40), unmet need (0.35), and genetic-evidence / druggability (0.25) — on cited 1–5 tiers. The autism signature shows up immediately in the unmet-need tier: there is no approved pharmacology for the core features of autism — the only licensed agents target the irritability adjunct, not the social-communication core or the E/I set-point itself — so, unlike schizophrenia (where the established D2 route drove the D2 receptor's unmet need low), no autism gene has its unmet need lowered by an established core route. The unmet-need floor here is higher than in any prior levers chapter; need is uniformly high. The substantive consequence is that the L2 inhibitory-restore route surfaces as the cleanest actionable direction: the GABA-A β3 subunit GABRB3 tops the actionable ranking, and the top actionable set is dominated by the L2 genes (GABRB3, GABRA5, KCNQ3, GABRA2) — not because L2 scores highest in the abstract, but because the high-scoring L1 excitatory genes are flagged not actionable by the generic sign: SCN2A and GRIN2B because their gain/loss sign-subtlety makes ‘reduce excitation’ a non-clean direction, and CACNA1C because it is a cross-disorder calcium set-point rather than an autism-selective handle. Filtering to clean directions lifts the inhibitory-restore lever to the top: restoring inhibition raises the fold regardless of variant direction, whereas reducing excitation runs straight into the sign problem. The single L3 node SLC6A4 is flagged not-actionable (non-monotone) and ranks last.

Crucially, the γ read is carried alongside each target as structural context but is never folded into the score — and the result is a clean demonstration of the firewall: the priority ranking and the γ / |h_sp| ranking are decoupled. The stiffest lever read in the set, the serotonin transporter SLC6A4 (|h_sp| ≈ 0.72), sits at the bottom of priority and is non-actionable; the top-priority target GABRB3 has only a mid-range read (|h_sp| ≈ 0.64). If promoter stiffness drove the ranking, neither could sit where it does — and across the full sixteen-gene set the decoupling is even starker, since the stiffest reads are the out-of-reach SHANK3 and RELN while the softest are the actionable SCN2A and CACNA1C. That decoupling is the firewall made visible, and it must be stated once more in full: the promoter |h_sp| is a gene's own switch stiffness, and it is never equated with the §18 network over-excitation threshold on R, nor with a receptor occupancy, a synaptic level, a compound's potency, a dose, an in-vivo selectivity, or any clinical effect. A fail-closed forbidden-claim scanner guards the whole package, and for autism it is the strictest in the series: beyond the usual dose / efficacy-as-fact / safety-as-fact / synthesis classes, it adds an autism quackery class — it rejects chelation, MMS / miracle-mineral / chlorine-dioxide bleach-protocol vocabulary outright, because that industry has harmed autistic children — and a normalise-framing class that rejects cure-autism, reverse-autism, and make-normal language, because autism is a neurodevelopmental difference, not only a deficit, and the map must never emit a disrespecting frame. Each class carries a planted self-test that must fire on its own bait, failing the build if it ever does not. This module reproduces bit-for-bit with the engine byte-unchanged.

Everything here is an in-silico reading of promoter sequence and a frame for organising targets, not a clinical measure, a diagnosis, or a prescription. The model asserts mechanism directions and target placements — an over-excitation fold can be raised by reducing inward current or restoring inhibition; these ten genes populate the levers; autism loads them L1-dominantly with a sparse L3; the map reaches the excitability axis only; the output and wiring axes are named and out of reach, the wiring axis provably so; restoring inhibition is the cleanest actionable direction because the excitatory channels are gain/loss sign-subtle — and nothing about which agent acts on any lever, at what dose, in whom, whether any real compound changes anyone's traits, or that anyone should change anything. The agents named as directions are illustrations of a sign, never a recommendation, and the recorded caveats (the gain/loss sign-subtlety of SCN2A and GRIN2B, the seizure edge of an over-pushed T lever, the non-monotone serotonergic node) are there precisely because a generic lever sign is not a clinical direction. Real autism is heterogeneous and is three axes, not one — excitability, output, and wiring — and that heterogeneity, including the two axes this lever map explicitly cannot reach, is locked. Autism is a neurodevelopmental difference, not only a deficit; this chapter ranks where a clean unmet mechanistic direction exists, and it does not assert that any direction treats, normalises, or cures autism. A promoter read and a lever assignment are mechanism boundaries, not a claim about the felt quality of an autistic mind (Axis-A firewall — consciousness_claim = 0, the hard problem stays open). This is not medical advice, not a diagnosis, not a treatment protocol, and not a cure. medium_efficacy_tested = 0; targets ranked, never drugs or doses.