Receptor-Subtype + Interaction-Effects Layer (model <-> intervention)

Created: 2026-06-12 Status: working abstraction layer (not a claim cluster; no claims.yaml entries created here) Depends on (reads): MECH-083..088 (four-plane neuromodulatory model), MECH-006/048/085, ARC-005, psychiatric_failure_modes.md, psychiatric_failure_axes.md (ARC-086) Feeds: evidence/planning/pharmacological_predictions.v1.json (PHARM-015..PHARM-022 + PHARM-001/010 orexin retrofit)

HONESTY RAIL (read first). REE is a research-stage computational specification, not a deployed AI, not a validated disease model, and not a clinical decision tool. Everything below is a set of testable hypotheses and shared vocabulary, not treatment guidance. The repurposing candidates named here are predictions to be adjudicated against clinical literature (RCTs, naturalistic cohorts), exactly like the rest of the PHARM registry. No patient should be treated on the basis of this document. Receptor pharmacology is the place REE has historically been “philosophy-right / mechanism-wrong” (see memory/feedback_biology_before_definitions): elegance is not evidence.

This layer cannot produce treatment evidence — only treatment-research directions. REE is never a link in the evidence chain. What it does is nominate which existing-drug trial is worth running and state the falsification condition in advance, so the trial is informative either way. The evidence — if it ever exists — is the trial’s, not REE’s. A receptor match plus a coherent mechanism story is a reason to test, never a reason to prescribe or to claim an effect. Every “predicted_effect” below is a hypothesis about what such a trial would find, not a finding.

1. Why this layer exists (the missing rung)

REE already has two mappings that sit on either side of a gap:

  REE mechanism claims              [drug-CLASS] PHARM registry
  (MECH-083..088 four planes,       (PHARM-001..014: DORA, SSRI,
   MECH-006/048/085, SD-036/037,     anticholinergic, GABAergic,
   psychiatric_failure_modes.md)     melatonergic, ...)
        |                                         ^
        |                                         |
        +------------- ??? -----------------------+
                  no receptor-resolution rung

The existing PHARM entries jump from a mechanism (“damp the SD-037 override signal”) straight to a drug class (“DORA”). That skips the rung where most modern psychiatric drug innovation — and most repurposing opportunity — actually lives: which receptor subtype, on which projection, with which interaction effects. Two drugs in the same coarse class (e.g. “serotonergic”) can act on opposite REE planes; one receptor (e.g. M4) can modulate a different plane than its parent transmitter’s headline role. The class label hides this.

This layer inserts that rung. It does three things the class layer cannot:

  1. Resolves an existing internal conflict by receptor subtype. The MECH-006 vs MECH-085 serotonin conflict (does 5-HT set temporal/rollout depth or not?) is flagged in neuromodulatory_control_planes.md with the receptor-subtype split named as the pending reconciliation. This document executes that reconciliation (Section 3A).
  2. Names interaction effects (M1 plasticity-gain x M4 dopamine-selection; mu/kappa opposing commitment-entropy) that are invisible at transmitter granularity (Section 3).
  3. Surfaces approved-drug repurposing candidates whose receptor profile matches a REE plane failure but whose labelled indication does not yet cover it — the highest-yield, lowest-cost- of-evidence interventions, especially for public health systems (Section 5).

2. The resolution table (subtype -> plane -> failure -> approved drug)

Each row reads: a receptor subtype, the REE plane/mechanism it carries, the failure signature when it is mis-tuned, an approved drug whose receptor action matches, and that drug’s current label status. “Repurposing gap” = the REE-predicted indication is not the drug’s current label.

Receptor subtype (projection) REE plane / mechanism Failure signature Approved drug (action) Current label Repurposing gap PHARM
5-HT2A (dorsal raphe -> L5 cortical pyramidal) Narrative-level representational collapse / exclusivity & top-level precision (MECH-006 rho-exclusivity; MECH-085 narrative level) Aberrant high-level salience, “narrative lock” / delusion-like over-precision (IMPL-005 #3 high alpha_theta; MECH-088 psychosis) Pimavanserin (5-HT2A inverse agonist) Parkinson’s disease psychosis (2016) Adjunctive MDD, primary psychosis (trialled, not labelled) PHARM-015
5-HT1A (median raphe -> hippocampal CA1/DG; + somatodendritic autoreceptor) Hippocampal-terrain basin geometry — aversive attractor wall steepness (MECH-085 terrain level) Deep aversive basins, fear-consistent rollout capture (anxiety; MECH-088 PTSD terrain component) Buspirone (5-HT1A partial agonist) Generalised anxiety disorder Augmentation in anhedonic/terrain-locked depression; PTSD terrain component PHARM-016
M1 (cortical/hippocampal muscarinic) ACh meta-plane: plasticity gain on bottom-up encoding (MECH-083) Low plasticity-gain -> new-context encoding failure / confirmation bias; chronically high -> map instability Xanomeline/trospium (M1/M4 agonist) Schizophrenia (2024, Cobenfy) MCI / AD cognition, AD psychosis (trialled, expanding) PHARM-017
M4 (striatal muscarinic) INTERACTION: dampens dopamine trajectory-selection gain (MECH-086) without D2 blockade Aberrant salience / spurious attractor selection (MECH-088 psychosis selection arm) Xanomeline/trospium (M1/M4 agonist) Schizophrenia (2024) First non-D2 antipsychotic — the receptor-resolved alternative to D2 antagonism PHARM-017
mu-opioid (mesolimbic/PAG) Commitment stabiliser — reduces policy-competition entropy (MECH-048; opioid thought 2026-02-11) Loss -> failure to consolidate selected commitments; excess -> rigid absorption Buprenorphine (mu partial agonist) Opioid dependence, pain — (partial; see kappa row) PHARM-018
kappa-opioid (dynorphin system) INTERACTION: commitment destabiliser — raises policy entropy, dysphoria/aversive state Pathological elevation -> anhedonia, dysphoria, aversive-state lock, suicidality Buprenorphine (kappa antagonist) (not labelled for this) Anti-anhedonic / anti-suicidal adjunct in TRD (trialled, not labelled) PHARM-018
NMDA-R (GluN2B; cortical pyramidal + fast-spiking GABAergic interneurons) RAPID map-geometry remodelling / aversive-basin flattening — the fast analogue of MECH-077 psychedelic remapping; deforms MECH-076 terrain topology / MECH-085 basin depth on an hours timescale, distinct from the 2-4 week monoaminergic route Rigid deep aversive basins resistant to slow monoaminergic remapping (treatment-resistant depression; acute suicidality) Esketamine (NMDA antagonist) / ketamine TRD + MDD with suicidal ideation (esketamine, 2019); racemic ketamine off-label Racemic ketamine (off-label) for the same rapid-remodelling indication — the major repurposing story PHARM-019
GABA-A alpha-2/3 (limbic / spinal interneurons) Harm-stream cross-stream decay regulator (SD-036) — anxiolysis as restored decay of pinned harm streams, without alpha-1 sedation Slowed harm-stream decay -> anxiety / harm-stream lock (catatonia subtype II, V3-EXQ-471) alpha-2/3-selective PAM (MK-0777/TPA023; darigabat) None approved (multiple Phase 2 GAD/panic; not converted) Non-sedating anxiolytic / harm-stream decay — the receptor-resolved subunit that carries anxiolysis PHARM-020
GABA-A alpha-1 (thalamocortical) Sedation / sleep-onset plane — hypnotic drive that bypasses homeostatic sleep pressure and does NOT preserve SWS architecture (INV-048) alpha-1 sedation substitutes for, rather than restores, SWS-dependent consolidation (dementia-pipeline cost, PHARM-009) Zolpidem (alpha-1-preferring PAM) Insomnia (approved) — (dissociation control: alpha-1 = sedation, NOT the anxiolytic-decay plane) PHARM-021
D2 (dorsal / associative striatum) Trajectory-selection GAIN plane (MECH-086 headline) — which rollout wins, how fast Excess -> premature commitment / aberrant salience; loss -> indecision D2 antagonists / agonists (antipsychotics; ropinirole) Schizophrenia / Parkinson’s (approved) — (headline selection arm, already labelled) PHARM-022
D3 (ventral striatum / nucleus accumbens, limbic) INTERACTION: anti-anhedonic incentive-salience arm of the selection plane (MECH-086) — reward-motivation weighting dissociable from D2 motor-selection gain D3 hypofunction -> anhedonia / amotivation (depression reward arm; MECH-088 DA selection failure) Pramipexole (D3-preferring agonist) Parkinson’s disease, restless legs (approved) Anti-anhedonic augmentation in depression (trialled, not labelled) PHARM-022
OX2R (lateral-hypothalamic orexin -> TMN / LC / raphe wake drive) SD-037 hyperarousal override / wake-promoting drive on the substrate (schema-repair-starvation plane) Tonic over-drive -> hyperarousal insomnia, schema-repair starvation (SD-037) DORA (suvorexant / lemborexant / daridorexant; OX2R-dominant for sleep) Insomnia (approved) — (cross-link: PHARM-001/010 now receptor-resolved at the OX1R/OX2R rung) PHARM-001/010

3. Interaction effects (the part class-granularity hides)

3A. Serotonin: receptor subtype resolves the MECH-006 / MECH-085 conflict

neuromodulatory_control_planes.md carries an unresolved conflict: MECH-006 says serotonin does not select temporal/rollout depth; MECH-085 says it does. The file names the resolution path — “different raphe nuclei and receptor subtypes re-use the same motif at different hierarchical levels” — but defers it pending a literature pull.

Resolution (receptor-resolved): the two claims describe the same serotonergic motif (modulating resolution pressure / exclusivity) applied at two levels by two receptor populations:

  • 5-HT2A on cortical L5 pyramidal cells (dorsal raphe) carries MECH-006’s account: narrative-/ cognitive-level representational exclusivity and top-level precision. 5-HT2A agonism loosens high-level priors (the psychedelic remapping of MECH-077); 5-HT2A inverse agonism (pimavanserin) tightens aberrant high-level salience. This is the “global tau at the E-stack level” that MECH-006 correctly says serotonin does not set — 5-HT2A modulates exclusivity, not depth.
  • 5-HT1A on hippocampal CA1/DG (median raphe) carries MECH-085’s terrain account: aversive attractor basin depth and local rollout geometry. This is “hippocampal-terrain-level depth,” a different construct from global tau.

So MECH-006’s prohibition (“does NOT select temporal depth tau”) and MECH-085’s rollout-depth claim are both correct at their own receptor/level — the apparent conflict was a granularity artefact. This is a concrete worked example of why the receptor rung is load-bearing: without it, the two claims look contradictory; with it, they are complementary.

3B. Muscarinic: M1 (plasticity) x M4 (dopamine selection) — one drug, two planes

Xanomeline is an M1/M4 agonist. The REE-relevant point is that its two targets sit on two different planes of MECH-087’s hierarchy:

  • M1 acts on the ACh meta-plane (MECH-083) — plasticity gain on encoding.
  • M4 acts downstream, dampening dopamine trajectory-selection gain (MECH-086) — this is why xanomeline is antipsychotic without D2 blockade. It attacks MECH-088’s psychosis “selection” arm (dopamine stamping salience onto noise) from the muscarinic side rather than the D2 side.

This is the clearest interaction-effect example: a single molecule whose therapeutic action is the joint effect of an upstream-plane action (M1) and a downstream-plane action (M4). The class label “muscarinic agonist” tells you neither plane.

3C. Opioid: mu / kappa as opposing poles on the commitment-entropy axis

The opioid thought (2026-02-11, processed into MECH-048) frames mu-opioid as a commitment stabiliser that reduces internal policy competition — “reduced urge to switch.” Its mirror is the kappa/dynorphin system, which raises policy entropy and carries dysphoric/aversive state. REE reads anhedonic, dysphoric, aversive-state-locked depression as pathological kappa-driven commitment destabilisation (nothing holds; the aversive attractor wins by default).

Buprenorphine is a mu partial agonist + kappa antagonist. The kappa antagonism is the architecturally interesting arm: it should normalise pathologically elevated policy-destabilisation, predicting an anti-anhedonic / anti-suicidal effect distinct from monoaminergic antidepressants (which act on the serotonin terrain and dopamine selection planes, not the opioid commitment-entropy axis). The MDD trial record (Section 5) is consistent with this.

3D. NMDA / glutamate: rapid map-geometry remodelling on a different timescale

The serotonin terrain account (MECH-085) and the structural-deformation account (MECH-076) both make map-geometry change slow — the 2-4 week SSRI onset is, in REE terms, the time it takes to re-grade a basin via reconsolidation cycles. MECH-077 already names psychedelic-assisted therapy as the exception: “transient attractor flattening enabling large-scale map restructuring.” NMDA-receptor antagonism (ketamine / esketamine) is the other fast route to the same construct, and it is the one with a marketed product.

The REE reading is mechanistic, not a re-label of “antidepressant”:

  • Ketamine blocks NMDA receptors preferentially on fast-spiking GABAergic interneurons, producing a net disinhibition / glutamate surge (AMPA-throughput, synaptogenesis). In REE terms that is a transient drop in the rigidity of the hippocampal-terrain attractor walls — a fast flattening of the MECH-085 aversive basin and a window in which MECH-076 terrain topology can be re-graded in hours rather than weeks. This is why the architecture predicts the effect on the terrain axis (basin depth, rollout capture), not on the dopamine-selection axis (PHARM-022) and not on the 5-HT2A narrative-exclusivity axis (PHARM-015).
  • The distinctive, falsifiable prediction is therefore timescale + axis, not just efficacy: a rapid (hours) collapse of aversive-basin capture that a monoaminergic agent cannot produce on that timescale, and that — crucially — should be strongest where the basin is deepest and most rigid (treatment-resistant, high-suicidality presentations), exactly the population where esketamine carries its label. If ketamine’s benefit were uniform across basin depth, or fully reproduced by accelerating a monoaminergic agent, the “fast-remodelling-of-a-rigid-terrain” reading would be wrong.

This is the highest-profile repurposing line in the layer: esketamine is approved (TRD / MDD with suicidal ideation) but racemic ketamine — the cheaper, generic, off-label molecule — is the same mechanism, and the cost-of-evidence gap between them is the public-systems story (Section 5).

3E. GABA-A: alpha-1 (sedation/sleep) vs alpha-2/3 (anxiolysis/harm-decay) are two planes

PHARM-004 and PHARM-009 treat GABAergic agents at the class level: PHARM-004 reads benzodiazepine relief of catatonia as augmenting the SD-036 cross-stream decay regulator; PHARM-009 reads chronic hypnotic use as SWS-architecture disruption (INV-048). The subunit rung resolves these into two different REE planes carried by two different alpha-subunits of the same receptor:

  • alpha-2/3-containing GABA-A (limbic, spinal) carries the anxiolytic / harm-stream-decay action (SD-036). Restoring decay of pinned harm streams is anxiolysis in REE terms — the mode flip from ‘avoid’ back to goal-seeking that SD-036 predicts. The architecturally clean prediction is that an alpha-2/3-selective positive allosteric modulator should produce anxiolysis without sedation and without the SWS-architecture cost.
  • alpha-1-containing GABA-A (thalamocortical) carries the sedation / sleep-onset action — and, per PHARM-009, the SWS-architecture disruption. The hypnotic effect substitutes for homeostatic sleep pressure rather than restoring SWS-dependent consolidation.

So the non-selective benzodiazepine’s two headline effects — anxiolysis and sedation — sit on two different REE planes, and the subunit split predicts they are dissociable (this is the GABAergic analogue of the M1/M4 two-plane story in 3B). The clinical record is consistent with the difficulty of the dissociation, not yet its success: alpha-2/3-selective anxioselective compounds have been trialled repeatedly (Merck’s MK-0777/TPA023 across five Phase 2 GAD studies; AbbVie’s darigabat in panic disorder) but none is approved — the efficacy/tolerability separation has been hard to realise. REE’s contribution here is the prediction that the dissociation is real and worth the engineering, with zolpidem (alpha-1-preferring) as the sedation-side control.

3F. Dopamine: D2 (selection gain) vs D3 (anti-anhedonic incentive) resolve MECH-086 by subtype

MECH-086 makes dopamine the trajectory-selection gain plane — “which rollout wins and how quickly.” Its failure modes are explicitly two-sided: “too little -> indecision, flattened landscape, anhedonia; too much -> premature commitment, aberrant salience.” The receptor rung resolves these onto two receptor subtypes with different anatomy:

  • D2 (dorsal / associative striatum) carries the selection-gain headline — the premature-commitment / aberrant-salience arm that D2 antagonists (antipsychotics) damp and D2 agonists push. This is MECH-086 as written and the MECH-088 psychosis “selection” arm.
  • D3 (ventral striatum / nucleus accumbens, limbic) carries the anti-anhedonic incentive-salience arm — reward-motivation weighting on the selection landscape, dissociable from D2 motor-selection. REE reads anhedonic / amotivational depression (MECH-088’s “DA selection failure” component) as a D3-resolved hypofunction of the incentive arm, distinct from the 5-HT2A narrative axis (PHARM-015), the 5-HT1A terrain axis (PHARM-016) and the kappa commitment-entropy axis (PHARM-018).

Pramipexole is a D3-preferring agonist (its D3:D2 affinity ratio is the relevant pharmacology). The architecturally distinctive prediction is that its anti-anhedonic signal should track the D3 (limbic incentive) arm and be dissociable from a pure D2 agonist’s motor effect — and that it acts on the reward-weighting arm of MECH-086, not on the serotonin terrain. The trial record is the strongest of the four extensions for an anhedonia-specific endpoint: a dedicated anhedonia-augmentation Phase 2 (NCT04121091) and a ventrostriatal-reward mechanistic Phase 4 (NCT02033369) both target exactly this arm. Pramipexole is approved for Parkinson’s disease and restless legs — the depression indication is off-label, the repurposing yield.

3G. Orexin: OX1R/OX2R is the receptor rung under the existing DORA entries

PHARM-001 and PHARM-010 already map the DORA class (suvorexant / lemborexant / daridorexant) onto the SD-037 hyperarousal-override substrate and the INV-048 sleep-architecture-preservation axis. They sit at the class level; the receptor rung underneath them is the OX1R / OX2R split. The sleep-promoting action is OX2R-dominant (OX2R drives the histaminergic TMN wake system; OX1R is more noradrenergic/reward-coupled), which is why DORAs and the OX2R-selective seltorexant are sleep agents. Adding the receptor_resolution field to PHARM-001/010 makes that rung machine-visible without changing the predictions: the override that DORAs damp is carried predominantly through OX2R onto the SD-037 wake-drive plane. This is a cross-link, not a new repurposing candidate — DORAs are already approved for insomnia — but it brings the two pre-existing orexin entries onto the same receptor-resolved footing as PHARM-015..022.

4. Substrate-modelling probe: what V3 can and cannot test

The user’s second question — can the current V3 substrate model these failure modes? — has a clean, honest answer with a sharp boundary.

V3 CAN perturb plane-level failures (existing ablation knobs, confirmed in ree-v3/):

REE plane failure V3 knob / ablation that reproduces it Existing script family
Dopamine selection flattening -> anhedonia/indecision E3 selection-gain / commitment threshold v3_exq_*_monostrategy_*, *_zgoal_monostrategy_falsifier
Harm-stream lock (catatonia II analog) harm-stream weights / cross-stream decay v3_exq_533_mech102_harm_stream_ablation, *_508_arc033_e2_harm_s_*
Consolidation / sleep-pipeline degradation (dementia analog) sleep-phase ablation, consolidation ablation v3_exq_242_sd017_sleep_phase_ablation, *_pharm_sleep_disruption_equivalence
Commitment-gate paralysis (catatonia I / OCD analog) closure-operator / commitment gate v3_exq_*_closure_*, SD-034 cluster
Drive disruption drive_weight v3_exq_238_sd012_drive_weight_ablation

The key already-registered testable prediction is MECH-087’s cross-plane non-rescue: degrade the serotonin (terrain) axis and the resulting trajectory pathology should NOT be rescued by increasing dopamine gain; degrade the dopamine axis and it should. That is a substrate-tractable experiment and the natural next probe (see Section 7 — flagged for /queue-experiment, not run here).

V3 CANNOT test receptor-subtype resolution. The substrate has no 5-HT1A-vs-2A split, no muscarinic M1/M4 distinction, and no explicit opioid policy-entropy term. The receptor subtype distinctions are therefore out_of_domain for V3 (consistent with the existing out_of_domain epistemic category for pharmacology in REE_assembly/CLAUDE.md): they are adjudicated against clinical literature, not V3 runs. Insight from the probe: the substrate can validate the plane-level dissociation logic that the receptor layer rests on (MECH-087), but the receptor rung itself is a clinical-literature object. This is the right division of labour — and it tells us the single highest-value V3 experiment to license the whole layer is the MECH-087 non-rescue test.

5. Repurposing candidates and their evidence (MCP-validated, 2026-06-12)

Trial counts and NCT IDs below were pulled live from ClinicalTrials.gov; receptor pharmacology from ChEMBL. “Approved-but-off-this-indication” is the repurposing-yield flag.

  • Pimavanserin (5-HT2A inverse agonist). Approved 2016 for Parkinson’s disease psychosis (ChEMBL CHEMBL2448613; USAN stem -anserin = 5-HT2 antagonist). ACADIA’s CLARITY programme ran Phase 2 (NCT03018340) and Phase 3 adjunctive-MDD (NCT03968159, completed) — the MDD indication was trialled but not approved. A Mount Sinai trial (NCT06592833, recruiting) uses pimavanserin to block 5-HT2A and dissect psilocybin’s mechanism — a ready-made dissociation probe for the Section 3A level-split. Repurposing gap: adjunctive MDD / primary psychosis.
  • Buprenorphine (mu partial + kappa antagonist). 27 depression trials on ClinicalTrials.gov, including Stanford anti-suicidal Phase 3 (NCT04116528, completed), a French anti-suicidal Phase 3 (NCT03646058, recruiting, n=180), late-life TRD (NCT01071538), and the ALKS-5461 (buprenorphine+samidorphan, kappa-antagonism unmasked) MDD programme (NCT01381107). Approved only for opioid dependence / pain. Repurposing gap: anti-anhedonic / anti-suicidal adjunct in treatment-resistant depression — high public-health salience given the suicidality endpoint.
  • Xanomeline/trospium (M1/M4 agonist, Cobenfy). 40 trials; approved 2024 for schizophrenia, now expanding to Alzheimer’s psychosis (ADEPT-3, NCT05980949), AD agitation (NCT07011732), and schizophrenia cognition (NCT07084831). The first non-D2 antipsychotic — the live proof that the receptor-subtype-resolved mechanism is where the field is moving. Repurposing gap: MCI/AD cognition, AD psychosis (actively closing).
  • Buspirone (5-HT1A partial agonist). Approved for GAD. REE-predicted extension: terrain-basin flattening as augmentation in anhedonic/terrain-locked depression and the PTSD terrain component — distinct from the 5-HT2A (pimavanserin) narrative axis. Repurposing gap: terrain-component augmentation.
  • Esketamine / ketamine (NMDA antagonist). Esketamine (Spravato, ChEMBL CHEMBL2364609) is FDA- approved (2019, max_phase 4) for treatment-resistant depression and for MDD with acute suicidal ideation (the MDSI label rests on Janssen Phase 3 NCT03039192, n=226, completed; the adolescent MDSI Phase 3 NCT07227454 is now recruiting). Racemic ketamine is off-label with a deep trial record — 140 TRD trials on ClinicalTrials.gov including Janssen Phase 2 NCT01627782 and MGH’s suicidal-ideation infusion work (NCT01582945; long-term maintenance NCT05450432). Repurposing gap: racemic ketamine for the same rapid-remodelling indication — the same NMDA mechanism at a fraction of esketamine’s cost, the single highest-yield line for a public payer. (Mechanism note: the ChEMBL salt-level mechanism record is sparse; NMDA-receptor antagonism is the established, textbook pharmacology and is cited as such, not as an MCP-derived novelty.)
  • Pramipexole (D3-preferring agonist, Mirapex; ChEMBL CHEMBL3182733). Approved 1997 (max_phase 4) for Parkinson’s disease and restless legs. 24 depression trials, with an anhedonia-specific signal: a dedicated anhedonia-augmentation Phase 2 (NCT04121091, Region Skane, completed), a ventrostriatal-reward-motivation mechanistic Phase 4 (NCT02033369, NYSPI), bipolar-depression augmentation (NCT00893841), a pramipexole+ondansetron MDD Phase 2a (NCT03642964, CTC-501, active), and inclusion in UCSF’s precision-MDD Phase 4 (NCT06580041). Repurposing gap: anti-anhedonic augmentation in depression — distinct from the D2 motor-selection arm and from monoaminergic agents. No black-box warning. (D3-preference is the established pharmacology; the depression indication is off-label.)
  • alpha-2/3-selective GABA-A PAMs (no approved agent yet). The subunit-selective anxioselective programme is real but unconverted: Merck’s MK-0777/TPA023 ran five Phase 2 GAD studies (NCT00539578 n=270; NCT00543920; NCT00546624 head-to-head vs lorazepam; NCT00703833), and AbbVie’s darigabat reached Phase 2 in panic disorder (NCT05941442, terminated 2025). None reached approval — the anxiolysis/sedation separation has been hard to realise in practice. The REE contribution is the prediction that the dissociation is architecturally real (alpha-2/3 = SD-036 harm-decay anxiolysis; alpha-1 = sedation/SWS-disruption, zolpidem as the control). This is a development-target prediction, not an approved-drug repurposing candidate — flagged honestly as lower-yield than the ketamine/pramipexole lines because there is no marketed alpha-2/3-selective molecule to repurpose.

Why this matters for public systems

The product here is a research direction, not a treatment claim. An already-approved drug carries a known safety profile, so the trial that would generate the evidence is cheaper and faster to run than for a new molecule — but REE only points to which trial is worth prioritising; the trial, not REE, produces the evidence. For a public payer (HSE included) the value is purely in better-prioritised research: a falsifiable, mechanism-motivated hypothesis about an off-label use is a stronger reason to fund a trial than an untethered hunch. It is not, and cannot become, a basis for prescribing ahead of that trial. The discipline below (falsification conditions, dissociation logic) is what makes a hypothesis worth a trial’s cost — it does not substitute for the trial.

6. Functional-link audit — new candidate links worth recording

Gap-mining the existing taxonomy against the receptor rung surfaced links not yet captured, each gated on a falsification condition (the rule: record only if testable).

  1. 5-HT2A inverse agonism <-> alpha_theta over-precision (IMPL-005 #3 “narrative lock”). The implementation-failure-mode doc names high alpha_theta as delusion-like narrative lock but gives no neuromodulator. Candidate link: 5-HT2A tone sets top-level precision; inverse agonism should reduce narrative-lock severity. Falsifier: pimavanserin does not reduce delusional conviction in primary psychosis beyond its established antipsychotic-adjunct effect on hallucinations.
  2. M4 muscarinic <-> dopamine-selection plane (MECH-086) without D2. Candidate link: aberrant salience can be corrected from the muscarinic side. Falsifier: xanomeline’s antipsychotic effect is abolished by selective M4 (not M1) knockdown only if the effect is M1-mediated — i.e. the prediction is that M4 is necessary; M1-only would falsify the selection-plane attribution.
  3. kappa-opioid <-> commitment-entropy axis as a depression subtype axis (ARC-086). Candidate link: kappa-driven destabilisation is a vulnerability axis distinct from the serotonin-terrain and dopamine-selection axes already in ARC-086. Falsifier: kappa antagonism (buprenorphine arm) produces the same response profile as an SSRI on anhedonia-specific endpoints (would collapse the axis into the existing serotonin axis).

  4. 5-HT1A autoreceptor desensitisation <-> SSRI onset delay reframed at the terrain level. The 2-4 week SSRI delay is attributed in MECH-085 to slow map-geometry change; the 5-HT1A somatodendritic autoreceptor desensitisation timecourse is a receptor-level mechanism for exactly that delay. Falsifier: 5-HT1A partial agonist augmentation (buspirone, pindolol) does not shorten antidepressant onset (the pindolol-augmentation literature is the existing partial test).

  5. NMDA antagonism <-> rapid terrain remodelling (MECH-076/077/085) on an hours timescale. The terrain account makes map-geometry change slow; MECH-077 already excepts psychedelic flattening. Candidate link: NMDA-antagonist disinhibition is the second fast-remodelling route, predicted to act on basin depth (terrain), not selection gain. Falsifier: ketamine’s antidepressant effect is uniform across baseline aversive-basin depth/rigidity (no greater effect in the most treatment-resistant, deepest-basin presentations), OR is fully reproduced by accelerating a monoaminergic agent — either would collapse the “fast-remodelling-of-a-rigid-terrain” reading into generic efficacy.
  6. D3 (vs D2) <-> the anhedonic/incentive arm of MECH-086 as a reward-weighting subtype. Candidate link: the “too little dopamine -> anhedonia” arm of MECH-086 is D3-resolved (limbic incentive salience), dissociable from the D2 motor-selection arm. Falsifier: a D3-preferring agonist (pramipexole) and a non-selective/D2 agonist produce indistinguishable effects on an anhedonia-specific (not motor, not global-mood) endpoint — which would collapse the D2/D3 subtype split of the selection plane.
  7. GABA-A alpha-2/3 (vs alpha-1) <-> SD-036 harm-decay anxiolysis as a subunit-resolved plane. Candidate link: anxiolysis (alpha-2/3, harm-stream decay) and sedation (alpha-1, sleep-onset/SWS cost) are separable REE planes carried by different subunits. Falsifier: an alpha-2/3-selective PAM that achieves clinically meaningful anxiolysis shows the same sedation/SWS-architecture liability as a non-selective benzodiazepine — which would mean the planes are not subunit-separable as predicted. (The repeated failure of alpha-2/3-selective agents to reach approval is a weak prior against easy separability, not yet a falsification — efficacy, not the dissociation per se, is where they have faltered.)

These are candidate links for psychiatric_failure_modes.md / the audit, not registered claims. Promotion to a claim requires the biology-first lit-pull (the standing rule).

7. Next steps (not executed here)

  • Substrate experiment (V3-tractable): MECH-087 cross-plane non-rescue test — degrade serotonin- terrain axis, confirm dopamine-gain does not rescue; degrade dopamine axis, confirm it does. Route via /queue-experiment (do not hand-write). This is the one experiment that licenses the plane- level dissociation logic the receptor layer rests on.
  • Lit-pulls (biology-first, before any claim registration):
    • DONE 2026-06-12 — strands 3A/3B/3C grounded (the original pilot PHARM-015..018): evidence/literature/targeted_review_receptor_subtype_layer/ (11 entries). 3A raphe projection specificity + the 5-HT2A-cortical / 5-HT1A-limbic distribution (Ren et al. 2019 eLife, DR/MR projection segregation; Weber & Andrade 2010, 5-HT2A on cortical L5 pyramidal; Akimova et al. 2009, 5-HT1A in anxiety; Wu et al. 2021, median-raphe 5-HT tunes hippocampal stability). 3B M4-striatal dopamine-selection damping + the non-D2 antipsychotic route (Foster et al. 2016 Neuron, M4->striatal dopamine via CB2; Kaul et al. 2024 EMERGENT-3, xanomeline phase 3; McKinzie & Bymaster 2012, reciprocal M4-dopamine). 3C kappa/dynorphin dysphoria as a commitment-entropy substrate (Land et al. 2008, dysphoria encoded by dynorphin/KOR; Bruchas et al. 2009, review; Krystal et al. 2020 Nat Med, KOR antagonism re-engages reward circuitry; Bari et al. 2025, KOR antagonism restores OPTIMAL PERSEVERATION — the closest empirical analogue of commitment stability). lit_conf raised on MECH-006/048/085/086/087/088 + ARC-086; exp_conf unchanged (all out_of_domain — grounds DESIGN + PHARM falsification, promotes nothing on the V3 substrate axis). The kappa-driven commitment-entropy ARC-086 axis is now lit-supported and PROPOSED for registration (proposal-first; NOT yet in claims.yaml — awaiting user decision, with the optimality + multi-component refinements from Bari et al. folded in).
    • PENDING — strands 3D/3E/3F/3G (the extension PHARM-019..022 + orexin retrofit): NMDA-antagonist disinhibition + rapid synaptogenesis as fast terrain remodelling (3D); GABA-A alpha-subunit anatomy (alpha-2/3 limbic anxiolysis vs alpha-1 thalamocortical sedation, 3E); D3-vs-D2 ventral/limbic incentive dissociation (3F); OX1R/OX2R wake-drive specificity (3G) — not yet pulled.
  • Registry: PHARM-015..018 (pilot) + PHARM-019..022 (this extension: NMDA/glutamate, GABA-A alpha-2/3, GABA-A alpha-1, D2/D3 dopamine) added; PHARM-001/010 (DORA) retro-fitted with the receptor_resolution field (OX1R/OX2R). All carry receptor_resolution so the rung is machine-visible; all are out_of_domain (clinical-literature-adjudicated). Future entries should carry the field by default.

Cross-references

  • docs/architecture/neuromodulatory_control_planes.md (MECH-083..088; the conflict this resolves)
  • docs/architecture/psychiatric_failure_modes.md (clinical mappings)
  • docs/architecture/psychiatric_failure_axes.md (ARC-086 vulnerability axes)
  • docs/thoughts/2026-02-11_opioid_receptors.md (mu/kappa commitment-entropy)
  • evidence/planning/pharmacological_predictions.v1.json (PHARM registry; PHARM-015..022 + orexin retrofit)

</content> </invoke>

REE is developed by Daniel Golden (Latent Fields). Apache 2.0.