Azure B fails to match methylene blue on working memory
STATUS_LABEL: NEGATIVE
Claim
On a pre-registered rodent working-memory task, Azure B alone fails to match parent methylene blue — WM Δ ≤70% of MB (or outside a pre-registered MB-equivalence band) — despite Azure B’s stronger MAO-A potency under identical enzyme conditions.
Why it matters
Popular MAOI framing: if MB WM is “mild MAOI,” the more potent demethyl metabolite Azure B should match or beat MB on the same task. N5 flagged the metabolite confound and asked selective MAO-A-i ≠ MB; it did not run Azure B vs MB on WM and is not proof of this failure. Soft-claims: discussion — Azure B∩memory soft-complete 0; behavioral mismatch remains UNKNOWN.
Mechanism sketch
Redox-vs-MAO assumption-kill: if WM tracks MAO-A potency, Azure B (≥6× stronger MAO-A than MB in vitro) should match or exceed MB; if WM tracks parent redox/structure, Azure B underperforms on the same WM task. Azure B is MB’s demethyl metabolite and a high-potency reversible MAO-A inhibitor (IC50 ≈11 nM vs MB ≈70 nM under identical conditions); MB is a potent reversible MAO-A inhibitor; low-dose MB improves discrimination / metabolic memory; redox-disabled N-acetyl MB loses mitochondrial ET / neuroprotection (not a declared WM endpoint). Competing caveats: PK / brain exposure may make Azure B underperform without killing the MAO story; soft-complete empty ≠ demonstrated failure.
Prior art
- PMID 22197611 / DOI 10.1016/j.taap.2011.12.005 — Azure B MAO-A IC50 ≈11 nM vs MB ≈70 nM under identical conditions (Petzer 2012).
- PMID 17721552 / DOI 10.1038/sj.bjp.0707430 / PMC2078225 — MB potent reversible MAO-A inhibitor; serotonin-toxicity prediction (Ramsay 2007).
- PMID 17428524 / PMC2040387 — low-dose MB improves discrimination learning / metabolic memory with COX framing (Gonzalez-Lima class) — no Azure B arm.
- PMID 21454572 / DOI 10.1074/jbc.M110.208447 / PMC3091255 — MB alternative electron transfer; N-acetyl redox-dead loses mitochondrial complex activity / OCR / neuroprotection (Wen 2011) — not WM.
Baseline claimed
MB working-memory benefit is explained by mild MAO-A inhibition, so the more potent demethyl metabolite Azure B should match or beat MB on the same WM task (or N5’s metabolite confound already settles Azure B WM).
Baseline measured
Azure B stronger MAO-A than MB LITERATURE (PMID 22197611). MB MAO-A potency LITERATURE (PMID 17721552). Gonzalez-Lima discrimination/metabolic memory LITERATURE without Azure B arm (PMID 17428524). Wen redox-dead kills mito ET/neuroprotection, not WM LITERATURE (PMID 21454572). Azure B alone vs MB on same WM UNKNOWN (soft-complete Azure B∩memory/cognition = 0). ≤70% MB-equivalence fail PREDICTION. N5 does not already prove this failure — metabolite confound only.
Actionable limitations
- Do not treat N5 as proving N6 — N5 asked selective MAO-A-i ≠ MB and listed Azure B as confound; it did not measure Azure B vs MB on WM.
- MAO IC50 ≠ brain WM PD — Petzer is enzyme inhibition without a cognition endpoint.
- PK / brain exposure — Azure B could underperform from distribution, not mechanism; soft-complete empty ≠ negative result.
Method
propose-assay
Prediction variables
| name | kind | assay | threshold |
|---|---|---|---|
| AzureB_stronger_MAO_A_than_MB | LITERATURE | Azure B vs MB MAO-A IC50 under identical conditions (Petzer 2012) | Azure B ≈11 nM vs MB ≈70 nM (~6×) (PMID 22197611) — no invented KD beyond cited IC50 |
| MB_MAO_A_potency | LITERATURE | Reversible MAO-A inhibition (Ramsay 2007) | potent MAO-A block; serotonin-toxicity class warning (PMID 17721552) |
| GonzalezLima_MB_discrimination_memory_no_AzureB_arm | LITERATURE | Low-dose MB discrimination / metabolic memory + COX framing | memory improve with metabolic enhancement; no Azure B comparator (PMID 17428524) |
| Wen_redox_dead_mitochondrial_not_WM | LITERATURE | N-acetyl-MB vs MB on mitochondrial complex activities / OCR / neuroprotection (Wen 2011) | redox center disabled → loses mitochondrial ET signal (PMID 21454572) — not a WM endpoint |
| same_WM_AzureB_vs_MB_panel | UNKNOWN | Same WM task — Azure B alone vs parent MB (± optional selective MAO-A-i), with occupancy/exposure readout | dedicated Azure B vs MB same-WM panel not found (soft-complete 0) — UNKNOWN |
| AzureB_fails_MB_equivalence_despite_stronger_MAO | PREDICTION | Pre-registered delayed alternation / Gonzalez-Lima-style discrimination; Azure B alone vs parent MB | Azure B WM Δ ≤70% of MB (or outside pre-registered MB-equivalence band); α/N pre-specified — do not invent observed Δ or dosing |
| competing_PK_exposure_explains_underperformance | PREDICTION | Same design; falsifier path if Azure B matches MB once brain exposure / MAO occupancy is matched | PK-matched equivalence allowed as pre-registered competing outcome |
Ablate
- remove same-task WM contrast of Azure B alone vs parent MB: Without the behavioral H2H, stronger MAO-A LITERATURE and MB memory LITERATURE cannot show Azure B fails to match MB on WM.
- remove stronger-MAO-A / MAO-comparable framing (Azure B vs MB enzyme potency anchor): Without the potency contrast, underperformance is just “Azure B weak,” not an assumption-kill of MAOI-explains-WM.
- remove N5≠proof + soft-complete-empty≠failure + PK-competing discipline: Treating N5 or empty PubMed as already deciding Azure B WM failure would invent a negative result left UNKNOWN.
Refute if
On the pre-registered WM task, Azure B alone matches parent MB within the MB-equivalence band (WM Δ >70% of MB or inside band) — especially under exposure/occupancy-matched conditions — so MAO-potency / metabolite-sufficiency for WM survives.
Ask a human
Lock WM task (delayed alternation vs discrimination vs OR), Azure B vs MB exposure/occupancy readout, and optional selective MAO-A-i arm before treating any MB cognitive signal as metabolite-MAO; do not treat N5 as already proving Azure B WM failure; no dosing.
Cheap IP
kill_cost_tier=1 · time_to_refute=4w · ip_class=redox-vs-MAO · vault_hold=no
Execution
execution_status=NOT_RUN · fto_status=UNKNOWN · related_prior=N5 (not proof of Azure B WM failure)