🧬 Aging Wiki

R40

40 min read

log/R40.md — Round 40 entries

Sub-file of log — see parent for index.

[2026-05-19] R40 close — effector proteins + dermal-matrix + wound-healing (12 pages)

Third round of the R38–R44 skin-aging coverage campaign. 12 atomic pages seeded + verified single-day. 24 agent invocations (12 seeders + 12 verifiers) + 1 propagation pass. All 12 pages flipped to verified: true (5 with partial scope for closed-access primaries).

Pages added (all 12)

Effector proteins (11): molecules/proteins/{mmp-1, mmp-3, mmp-9, timp-1, col1a1, col3a1, eln, fbn1, lox, kitlg, kit}.md

Process (1): processes/wound-healing.md

Major verifier findings

R40 verifier yield was the highest of any round in the campaign so far — ~70 substantive corrections across 12 pages including multiple direction inversions, DOI fabrications, and biology errors.

Critical biology corrections:

  • MMP-9 Tarantini 2021 direction inverted: seeder called it a “null result”; actual result is paradoxical worsening — MMP-9-KO mice show 77% ICH incidence vs 35% WT (p=0.017). MMP-9 is protective in this context, not harmful.
  • COL3A1 Col3a1-null mouse phenotype WRONG: my seeder brief asserted “embryonic lethal from vessel rupture” (from training memory); Liu 1997 PNAS (verifier downloaded + read) shows neonatally lethal (~95% within 48h of birth, NOT embryonic). ~10% survive to adulthood with ~6-month lifespan, dying from aortic/intestinal rupture.
  • Inomata 2009 primary model fabrication (caught earlier in R39) propagated forward to R40 framing: real model is Dct-lacZ, not K14-SCF.

DOI fabrications and misattributions:

  • MMP-9 two DOI errors: Liu 2022 1044877 → 1021129; Hu 2024 112110 → 112290 (wrong DOI resolved to unrelated DPP-IV/ulcerative colitis paper).
  • MMP-9 trial name WRONG: andecaliximab Phase 3 trial called “RAINBOW-2” → GAMMA-1 (RAINBOW-2 is a totally different ramucirumab/paclitaxel trial).
  • MMP-3 Sumsuzzman 2022 DOI off by one digit: 101730 → 101729 (101730 is an unrelated ischemic-stroke paper).
  • MMP-3 Mak 2025 DOI completed: incomplete 10.1016/j.brainres.2025 → full 10.1016/j.brainres.2025.149954.
  • TIMP-1 Nagase 2006 DOI WRONG: 10.1093/cvr/cvl248 → 10.1016/j.cardiores.2005.12.002 (the wrong DOI returns 404).
  • LOX Herchenhan 2015 DOI WRONG: 10.1074/jbc.M114.626523 (yeast CK2 paper) → 10.1074/jbc.M115.641670 (actual LOX/tendon paper; colliding DOI on different paper).
  • LOX Liu 2004 DOI WRONG: 10.1038/ng1391 → 10.1038/ng1297.
  • COL1A1 Shuster 1975 BUG-2 resolved: PMID 1220399 (Russian ENT paper) → PMID 1220811; DOI 10.1111/j.1365-2133.1975.tb05113.x.
  • COL3A1 Kong 2016 DOI resolved: confirmed 10.1111/jocd.12193.
  • ELN Shapiro 1991 elastin half-life source resolved: previously #gap/unsourced; 10.1172/JCI115204 (74-yr lung residence time; lung-specific, not directly aortic/dermal).
  • ELN Tassabehji 1998 exon number WRONG: “exon 30” → exon 32 (PMID 9580666 abstract confirmed).
  • Dong 2025 phantom reference resolved: surfaced by R39 Inomata verifier; PubMed returns ZERO results. Real paper is Mohri 2025 (PMID 41053225, Nat Cell Biol).

Quantitative / canonical-database fabrications by main-agent training memory:

  • MMP-1 mouse ortholog UniProt accessions WRONG: my brief asserted Q9JHH5 / Q8K157 (which resolve to CXCL11 chemokine + GALM galactose mutarotase); actual is Q9EPL5 / Q9EPL6 (Mmp1a / Mmp1b).
  • KIT Hirota 1998 94% misattribution: I said “94% of GIST have exon-11 mutations”; paper actually says 94% (46/49) = KIT expression frequency by IHC, NOT mutation rate (only 6/49 sequenced; ~60-70% mutation rate in larger subsequent series).
  • KIT signal peptide length: 21-aa → 25-aa (UniProt SIGNAL 1..25).
  • MMP-1 domain boundaries: catalytic 100-270 → 100-269; hemopexin start 293 → 270 (off by 23 residues).
  • TIMP-1 NTR domain boundary: 24-127 → 24-147 (off by 20 residues; UniProt P01033 feature table).
  • TIMP-1 MMP-2 framing: implied TIMP-1 doesn’t inhibit MMP-2; UniProt confirms it does — the distinction is TIMP-2 dominates MMP-2 activation via MT1-MMP ternary complex (because TIMP-1 doesn’t inhibit MT1-MMP/MMP-14, NOT because TIMP-1 doesn’t bind MMP-2).
  • TIMP-1 substrate list missing MMP-12 + MMP-16 (UniProt-confirmed substrates).
  • MMP-3 catalytic domain boundary: 100-266 → 100-272.
  • COL3A1 dermal proportion denominator WRONG: “6-9% of dry dermal weight” was the wrong denominator — Shin 2019 actually says “8-12% of total collagen” (R38 framing was also off).
  • COL3A1 I:III ratio specific values unsourced: “young ~5:1 → aged ~3:1” not in Shin 2019 or Fisher 2009; replaced with derivable estimate.
  • LOX Lox-/- lethality description: “survivors die from aortic aneurysm” was fabricated detail; Staiculescu 2017 reports perinatal death within hours of birth, no surviving cohort dying later.
  • LOX AAT10 Arg-298 overclaim removed (UniProt has no activity quantification for that variant).
  • LOX Tyr-355 cofactor precision: “topaquinone” → “2’,4’,5’-topaquinone” per UniProt exact annotation.
  • FBN1 ~3,000 pathogenic variants figure removed (unverifiable against current LOVD 1,681 / ClinVar 9,445 submissions).
  • FBN1 losartan human-trial framing substantially overstated: I said “in some human trials”; three largest RCTs (PHN n=608, Marfan-Sartan n=303, Spanish n=140) all found no significant losartan benefit over atenolol. Kang 2020 meta-analysis (PMID 31003918) added as correction citation.
  • FBN1 domain count source corrected: anchored to Zeigler 2021 verified PDF (47 EGF / 43 cbEGF / 7 TB) instead of not_oa Kielty 2005.
  • ELN solar elastosis Chen 2005 modality mischaracterised: I said “UV”; paper is actually “Heat Modulation” — infrared, NOT UV.
  • ELN Weihermann 2021 MMP12/LOX significance overstated: corrected to compartment-specific values (MMP12 only significant in ex-vivo photoexposed; LOX only in periocular blepharoplasty 1.64-fold).
  • Wound-healing PDGF isoform distinction clarified: becaplermin = PDGF-BB/PDGFRβ; Demaria 2014 mechanism = PDGF-AA/PDGFRα — important for preserving the senolytic-timing implication accurately.
  • KIT D4-D4 → D4-D5 homodimer contact (typo).
  • COL1A1 Varani 2006 n-values WRONG: “~9-10 donors” → n=6 in-tissue / n=26+37 isolates from 8 donors per group in-vitro. “~1%/yr rate” misattribution removed (paper reports 68% cross-sectional reduction, not annualized rate; rate is Shuster 1975 only).
  • MMP-3 sub-erythemogenic threshold misgeneralised: 0.01 MED threshold applies to MMP-1 per Fisher 1996 (NOT MMP-3); MMP-9 threshold is 0.1 MED.

Coppé 2008 SASP supplemental-vs-main-text qualification surfaced by BOTH MMP-1 + MMP-9 verifiers:

  • MMP-1 is in supplemental Table S2 (120-protein antibody array catalogue), NOT in main text or primary figures.
  • MMP-9 is NOT on the Coppé 2008 array at all — the antibody array focused on cytokines/chemokines/growth factors. The widespread “MMP-9 is a SASP component per Coppé 2008” claim in the literature is unsupported by that specific paper; MMP-9-in-SASP comes from independent gelatinase studies.
  • Propagated qualification to processes/replicative-senescence.md § Step 4 SASP acquisition.

KITLG / Imokawa 2019 review-vs-primary discrepancy on c-KIT mRNA:

  • Imokawa 2019 review claims c-KIT mRNA 2.14-fold elevated in solar lentigo, citing Hattori 2004.
  • R39-verified Hattori 2004 primary actually reports 1.16-fold NS. Page retains the verified primary value with discrepancy note.
  • Imokawa 2019 is widely cited; this is a propagating review-vs-primary error in the broader literature.

Propagation work completed

  • molecules/proteins/fbn1.md Chen 2005 framing corrected: “UV-exposed skin” + “UV-induced heat” → “infrared/heat” properly attributed (matching ELN page correction).
  • processes/replicative-senescence.md § Step 4: “MMP-3, MMP-9” → “multiple MMPs” with explicit Coppé 2008 supplemental-vs-main-text qualification + cross-link to verified mmp-9.md.
  • processes/sasp.md intro: existing language (“MMPs”) is general enough not to need correction; main-figure-vs-supplemental qualification documented on individual MMP pages instead.
  • ROADMAP.md R45 forward-queue: studies/dong-2025-msc-seno-differentiation.md → studies/mohri-2025-msc-seno-differentiation.md (phantom reference corrected to PMID 41053225).
  • Open Targets API up during MMP-1 verifier pass: max-druggability tier 1 (broad MMP inhibitors approved for non-aging indications); aging-context tier 3 retained per CLAUDE.md R26/R27 convention.

Propagation work confirmed not needed:

  • Sumsuzzman 2022 wrong DOI (101730): grep wiki-wide returned no further occurrences beyond mmp-3.md scope + log entries.
  • Tarantini 2021 ICH direction-inversion: 2 other pages (plasma-exchange.md + time-restricted-eating.md) reference different Tarantini-coauthored papers (Gulej 2026 / Natarajan 2026), not Tarantini 2021 ICH. No further propagation.
  • Tassabehji “exon 30” appears only on ELN page (corrected to exon 32) + log entries.
  • ~3,000 FBN1 variants figure appears only on FBN1 page itself (corrected).

Schema escalations surfaced

  • type: process druggability-tier convention — wound-healing seeder flagged: process with FDA-approved modulator for non-aging indication (becaplermin for diabetic foot ulcers). Verifier used tier 1 with inline rationale. The aging-context-tier distinction from R26/R27 protein/pathway convention should extend formally to type: process. Adding to next CLAUDE.md cleanup queue.
  • ELN under loss-of-proteostasis hallmark using López-Otín 2023 extended ECM-proteostasis framing — schema decision documented; verifier confirmed reasonable; no formal CLAUDE.md cleanup needed.

Forward queue surfaced by R40

  • Coppé 2008 study page seeding candidate — both MMP-1 and MMP-9 verifiers surfaced this as worth a dedicated extraction (cited from many pages but no dedicated study page; the supplemental-vs-main-text qualification deserves an authoritative anchor).
  • Studies seeding candidates: studies/varani-2006-collagen-aged-skin.md (R40 COL1A1 verifier resolved its DOI + read PDF; downloaded — natural next seed); studies/liu-1997-col3a1-knockout-neonatal-lethality.md (R40 COL3A1 verifier downloaded; resolves the previously gap-tagged knockout paper); studies/shapiro-1991-elastin-half-life-lung.md (R40 ELN verifier downloaded); studies/young-2024-kitl-msc-hsc-aging.md (R40 KITLG verifier downloaded from PMC); studies/zeigler-2021-marfan-pathophysiology.md (R40 FBN1 verifier downloaded); studies/hirota-1998-gist-kit-mutation.md (R40 KIT verifier resolved 94%-expression-not-mutation-rate misreading).
  • Imokawa 2019 vs Hattori 2004 review-vs-primary c-KIT discrepancy — worth a wiki-wide propagation check; review-derived numbers should be flagged where they contradict the R39-verified Hattori primary.

Round disposition

R40 closed cleanly. 12/12 pages flipped to verified: true. ROADMAP.md updated. Ready to dispatch R41 (retinoid cluster + class MOC; 7 pages).

[2026-05-19] verify — molecules/proteins/lox.md

Pages verified: 1

  • molecules/proteins/lox.md — flipped to verified: true (partial scope)

Corrections applied (8):

  1. Herchenhan 2015 DOI wrong: 10.1074/jbc.M114.626523 → 10.1074/jbc.M115.641670 (the archived path served a CK2/yeast paper with the wrong DOI; PubMed efetch of PMID 25979340 revealed the correct DOI).
  2. Herchenhan 2015 model wrong: “rat tail tendon cells” → “human tenocyte-derived tendon constructs” (the paper used human tenocytes to form tendon-like constructs, not rat cells).
  3. Liu 2004 DOI wrong: 10.1038/ng1391 → 10.1038/ng1297 (PMID 14745449; efetch confirmed correct DOI and expanded LOXL1 phenotype to include lung enlargement, loose skin, and vascular abnormalities beyond just pelvic organ prolapse).
  4. AAT10 Arg-298 overclaim removed: wiki said “causes loss of function” — UniProt does not annotate quantitative activity data for Met-298→Arg; corrected to note it is an AAT10-classified variant with no reported activity quantification. Variant also clarified as Met→Arg substitution.
  5. Tyr-355 precision improved: “topaquinone” → “2’,4’,5’-topaquinone” matching UniProt’s exact annotation.
  6. Lox⁻/⁻ lethality expanded: “cardiovascular failure” → “cardiovascular and pulmonary abnormalities” per Staiculescu 2017 PDF text; “within a few hours of birth” added; removed erroneous sentence “Survivors die from aortic aneurysm.”
  7. Cenizo 2006 PMID added (16842595) and LOXL-predominant age effect nuance added to body text and footnote; closed-access #gap/no-fulltext-access tag added.
  8. Moulin 2017 PMID added (27396912) and abstract-derived details expanded (ChIP, luciferase reporter, Origanum extract findings); closed-access #gap/no-fulltext-access tag added.

Additional footnote improvements:

  • Foote 2024, Yao 2022, Vachhani 2025, Staiculescu 2017: “locally pending download” removed; footnotes expanded with PDF-verified quantitative details.
  • Schmelzer 2020: “locally pending download” note removed (download completed).
  • #gap/unsourced in Limitations resolved and replaced with precise Cenizo 2006 citation framing.

Unverifiable claims (closed-access PDFs):

  • Cenizo 2006 (PMID 16842595) — abstract-level confirmation only; fibroblast sample counts, exact RT-PCR fold-changes, statistical details unverified.
  • Moulin 2017 (PMID 27396912) — abstract-level confirmation only; exact ChIP quantification, methylation percentages, LOXL1 secretion restoration data unverified.
  • Behmoaras 2008, Szauter 2005, Remus 2012 — not_oa; claims low-priority background; not re-checked.

Supersession candidates (R25): None. No post-2024 meta-analysis or large RCT on LOX/aging found. Vachhani 2025 is the current frontier for clinical LOX inhibition; it confirms the myelofibrosis indication context (not aging), consistent with wiki framing. literature-checked-through: 2026-05-19 already set in frontmatter.

Downstream propagation needed:

  • tissues/dermis.md — the #gap/unsourced LOX claim is now resolved via Cenizo 2006; dermis.md should reference [^cenizo2006] if it carries an independent claim about LOX activity decline.
  • hallmarks/loss-of-proteostasis.md and any page citing LOX family cross-linking as a proteostasis-adjacent process — no corrections needed, but verify the wikilink [[lox]] resolves correctly.

[2026-05-19] R40 verify — molecules/proteins/fbn1.md

Pages verified: 1

  • molecules/proteins/fbn1.md — flipped to verified: true (partial scope)

Corrections applied (7):

  1. Molecular weight qualifier removed: “~350 kDa after signal-peptide cleavage” → “350 kDa glycoprotein” — the qualifier “after signal-peptide cleavage” was not stated in sources (Zeigler 2021, Matt 2008 both simply say 350 kDa).
  2. Domain architecture citation corrected: [^kielty2005] → [^zeigler2021] as the verified source for domain counts (Kielty 2005 is not_oa). Added note documenting the discrepancy between Zeigler 2021 counts (43 cbEGF, 7 TB) and current UniProt flat-file annotation (40 cbEGF, 9 TB).
  3. Watson 1999 n and photodamage range corrected: “moderate-to-severely photoaged skin” → “both severely and minimally photoaged skin”; added subject demographics (n=16, 6M/10F, age 24–80y); added quantitative p-values (p<0.0001 severe; p<0.05 minimal); clarified that depletion was shown by confocal/IHC (not ISH which showed mRNA); ISH result clarified as 8/10 subjects with decreased FBN1 mRNA.
  4. Losartan human-trial framing corrected: “have been shown to reduce aortic root dilation in mouse Marfan models and in some human trials” → accurate statement that mouse data (Fbn1 C1039G/+) was strongly positive but multiple large human RCTs (PHN n=608; Marfan-Sartan n=303; Spanish trial n=140) found no significant benefit over atenolol. COMPARE (n=233) and Kang 2020 meta-analysis (8 RCTs, n=1,381) added as nuance.
  5. ~3,000-variant figure removed: Unsourced figure replaced with LOVD-verified counts (1,681 unique public variants; ClinVar 9,445 total submissions all classes). Precise pathogenic-only count flagged as #gap/unsourced pending ClinVar/ClinGen query.
  6. LTBP specificity qualified: “LTBP-1, -3, and -4” narrowed to “LTBP-1 confirmed; LTBP-3/-4 from Kielty 2005 (not_oa, unverified).”
  7. Not-downloadable sources tagged: Matt 2009 (Circulation, no OA URL), Chen 2005 (not_oa), Kielty 2005 (not_oa) all tagged #gap/no-fulltext-access in footnotes.

New citations added:

  • Kang 2020 (PMID 31003918) — 8-RCT meta-analysis, n=1,381 MFS patients; ARB non-inferior to BB; supports corrected losartan framing.
  • LOVD FBN1 gene page — accessed 2026-05-19; 1,681 unique public variants; supports corrected mutation-count claim.

Supersession candidates (R25): None identified. Kang 2020 meta-analysis is additive (updates losartan human framing) but not a supersession of the core FBN1 biology. PMID 36243587 (UVA/mitochondria/fibrillin-1 depletion mechanism, 2022) is a mechanistic elaboration consistent with the existing framing, not a supersession.

Unverifiable claims retained: Chen 2005 solar-elastosis paradox (epidermis vs. dermis heat modulation); Matt 2009 (losartan TGF-β suppression in human MFS patients); Kielty 2005 LTBP-3/-4 specificity — all tagged.

literature-checked-through: already set to 2026-05-19 by seeder; confirmed correct.

[2026-05-19] R40 verify — molecules/proteins/eln.md

Pages verified: 1

  • molecules/proteins/eln.md — flipped to verified: true

Corrections applied (7):

  1. Half-life unsourced gap resolved: #gap/unsourced cleared. Shapiro 1991 (J Clin Invest, doi:10.1172/JCI115204) identified as the primary radiocarbon source — lung parenchyma, mean carbon residence time 74 yr (95% CI 40–174 yr), n=14 post-mortem specimens. Ritz-Timme 2003 (Br J Dermatol) added as skin-specific corroboration (aspartate racemization, r=0.98, no numerical half-life).
  2. Tissue specificity of half-life corrected: Wiki body stated “~70-year half-life for human aortic and dermal elastin” → corrected to “74-yr mean carbon residence time in human lung parenchyma; skin extrapolation supported by Ritz-Timme 2003 but no direct radiocarbon measurement for skin/aorta.”
  3. Tassabehji 1998 exon number wrong: “exon 30 frameshift mutation” → “exon 32 frameshift mutation” (PubMed abstract PMID 9580666 confirmed). Mutation details added: replaces 37 C-terminal aa with 62-aa novel sequence.
  4. Chen 2005 radiation modality clarified: Title “Heat Modulation of Tropoelastin, Fibrillin-1, and MMP-12” — paper studies infrared/heat, not UV. Wiki body sections corrected from “UV irradiation” to “photoexposure/infrared irradiation” where referencing Chen 2005 specifically; no-mechanism reworded to include both UV and infrared stimuli.
  5. Weihermann 2021 MMP12 significance corrected: Wiki implied MMP12 was elevated in all photoexposed models — MMP12 elevation in irradiated reconstituted skin was 1.94-fold NON-SIGNIFICANT; significant MMP12 elevation was in ex vivo skin only (5.98-fold and 3.14-fold). Corrected in body text.
  6. Weihermann 2021 LOX elevation overstated: Wiki said “LOX mRNA is also elevated in photoexposed skin” — actual result: LOX elevated 1.64-fold in periocular (blepharoplasty) region only; auricular and reconstituted skin showed no significant LOX increase. Corrected.
  7. Weihermann 2021 irradiation modality corrected: Wiki said “UV-exposed reconstituted skin” → paper used a solar radiation simulator (not UV-only). Also: exon 26A fold-changes added (3.61× auricular, 2.58× periocular, 4.59× reconstituted skin).

Sources checked: Shapiro 1991 (PDF verified); Weihermann 2021 (PDF verified, gold OA); Tassabehji 1998 (PubMed abstract, DOI lookup failed); Chen 2005 (Crossref metadata, not_oa); Taddese 2008 (not_oa); Ritz-Timme 2003 (PubMed abstract, not_oa); UniProt P15502 (API confirmed).

Unverifiable: Taddese 2008 (not_oa) — MMP-12 cleavage site claims retained with #gap/no-fulltext-access; Chen 2005 (not_oa) — heat/infrared induction claims retained with gap tag; Tassabehji 1998 (archive failed) — exon 32 confirmed via PubMed abstract, full PDF not read; Lockhart 2020 (hybrid OA, download pending).

Downstream pages to check: phenotypes/skin-aging.md — may inherit the “UV” vs “infrared” distinction from Chen 2005; may state “exon 30” for Tassabehji. Main agent to propagate.

[2026-05-19] R40 verify — molecules/proteins/col3a1.md

Pages verified: 1

  • molecules/proteins/col3a1.md — flipped to verified: true

Corrections applied (7):

  1. Dermal dry-weight figure wrong: “~6–9% of dermal dry weight” → Shin 2019 actually states “8–12% of total collagen” (not dry weight). Corrected in tissue table and body.
  2. I:III ratio claim unsourced: “young ~5:1 → aged ~3:1” attributed to Shin 2019 — these specific values are NOT in that paper. Replaced with derivable I:III estimate (~7–9:1) from the 80–90% vs 8–12% proportions, plus #gap/needs-replication for the specific ratio-shift quantification.
  3. Col3a1 null lethality mischaracterized: wiki said “embryonic lethal” → Liu 1997 PNAS shows predominantly NEONATAL lethality (within 48 h post-birth); ~10% survived to adulthood (~6-month lifespan). Identity section and body corrected.
  4. Liu 1997 gap resolved: #gap/needs-canonical-id resolved. DOI 10.1073/pnas.94.5.1852 confirmed (Liu X, Wu H, Byrne M, Krane S, Jaenisch R · PNAS 1997 · 94:1852–1856). Footnote [^liu1997] added.
  5. Fisher 2009 type III attribution overstated: wiki implied type III fragmentation was separately studied in that paper — Fisher 2009 is entirely focused on type I collagen. Corrected + gap tag added for type III-specific contribution.
  6. Fisher 2009 n added: “n=not specified” → n=4–7 per assay (young: 21–30 yr; aged: >80 yr).
  7. Kong 2016 DOI resolved: Confirmed 10.1111/jocd.12193 (Kong R et al. J Cosmet Dermatol 2016;15(1):49–57).

Supersession (R25): PubMed search 2021–2026 for COL3A1 meta-analyses/RCTs returned 30 results; top hits are psychiatric genetics, pulmonary fibrosis, hand rejuvenation — none supersede aging-relevant claims. literature-checked-through: 2026-05-19 confirmed.

Downstream propagation needed: None — col3a1 is an atomic protein page with no downstream study-page inbound links propagating these specific claims.

[2026-05-19] R40 verify — molecules/proteins/kit.md

Pages verified: 1

  • molecules/proteins/kit.md — flipped to verified: true

Corrections applied (6):

  1. Signal peptide: “21-aa” → “25-aa” (UniProt P10721 SIGNAL 1..25).
  2. Ig-like domain per-domain residue ranges added (27–112, 121–205, 212–308, 317–410, 413–507).
  3. “D4–D4 receptor homodimer contact” → “D4–D5” (typo in original).
  4. Hirota 1998 footnote critically wrong: 94% figure is KIT expression (46/49 GISTs), not mutation rate; mutation sequencing was 5/6 GISTs (exon-11 JMD). Footnote corrected.
  5. Druggability-tier frontmatter: 1 → 2 (aging-context per R26/R27; body was already correct).
  6. Footnote access statuses: Hachiya 2009 + Mohri 2025 “pending” → “not_oa no-fulltext-access”; Gao 2024 download-failed noted; PMID 41053225 added to Mohri 2025.

Dong 2025 phantom resolved: No such paper in PubMed. Mohri 2025 (PMID 41053225) is correct citation. Supersession (R25): No aging-context RCTs/meta-analyses found. literature-checked-through: 2026-05-19 confirmed.

[2026-05-19] R40 verify — molecules/proteins/mmp-9.md

Pages verified: 1

  • molecules/proteins/mmp-9.md — flipped to verified: true

Corrections applied (8):

  1. Tarantini 2021 framing critically wrong: wiki said “null result / did not show delayed onset or reduced incidence” → corrected to paradoxical worsening: MMP-9 KO had 77% ICH incidence vs 35% WT (n=22 KO, n=17 WT; log-rank χ²=5.701, p=0.017). Body text, gap table, and footnote all corrected.
  2. Liu 2022 DOI wrong: 10.3389/fnagi.2022.1044877 → 10.3389/fnagi.2022.1021129 (confirmed via PubMed PMID 36337710).
  3. Hu 2024 DOI wrong: 10.1016/j.intimp.2024.112110 → 10.1016/j.intimp.2024.112290 (wrong DOI resolves to a completely different paper; correct paper confirmed via PubMed PMID 38796964).
  4. NGAL complex MW wrong: “~135 kDa” → “~125 kDa” per UniProt P14780 reference annotation.
  5. Andecaliximab trial name wrong: “RAINBOW-2” → “GAMMA-1” (Shah MA et al., JCO 2021, PMID 33577358); OS result corrected to 12.5 vs 11.8 months (ns). UC and RA trial stages corrected.
  6. Coppe 2008 SASP-MMP9 claim: PDF confirmed MMP-9 was NOT among the 120 proteins on the antibody array; body text and footnote corrected; gap tag added.
  7. Banner removed: ⚠️ auto-extraction banner removed.

Sources checked: Fisher 1996 via verified study page; Fisher 2009 via verified study page; Coppe 2008 full PDF read; Tarantini 2021 via PMC8599521; Liu 2022 / Hu 2024 DOIs confirmed via PubMed; UniProt P14780; Open Targets.

Supersession check (R25): no aging-specific RCT/meta-analysis found superseding existing framing (2022–2026). literature-checked-through: 2026-05-19 confirmed.

[2026-05-19] R40 verify — processes/wound-healing.md

Pages verified: 1

  • processes/wound-healing.md — flipped to verified: true (partial scope; Singer 1999 closed-access)

Corrections applied (5):

  1. Druggability-tier comment expanded: added aging-context rationale (wound-healing impairment is a major age-related clinical burden) and explicit note that becaplermin targets PDGFRβ (PDGF-BB isoform), distinct from the PDGF-AA/PDGFRα senescent-cell axis (Demaria 2014). Tier-1 designation confirmed correct under aging-context convention.
  2. Singer 1999 PMID added to footnote: correct PMID 10471461 confirmed via PubMed (NEJM 1999, vol 341, issue 10, pp 738–746). Prior footnote had DOI only.
  3. Tensile strength figure (~70–80%) tagged unverifiable: Singer 1999 is closed-access (not_oa in archive); #gap/no-fulltext-access tag added in table, remodeling body paragraph, and footnote.
  4. Gurtner 2008 footnote updated: added volume/page numbers (453:314–321), confirmed end-to-end verified 2026-05-19, added note that Gurtner uses three-stage model while four-phase model follows Singer 1999 convention.
  5. Banner removed; double blank line after frontmatter cleaned.

Sources checked: Demaria 2014 cross-checked against R39-verified study page (not re-read — study page is verified: true 2026-05-19); Gurtner 2008 full PDF read end-to-end (8 pp, local); Singer 1999 closed-access (not in archive — tensile strength figure unverifiable); Ueda 2026 DOI confirmed via Crossref; Miny 2026 DOI confirmed via Crossref; Numani 2026 DOI confirmed via Crossref.

Unverifiable: Singer 1999 (doi:10.1056/NEJM199909023411006, PMID 10471461) — closed-access (not_oa per archive); tensile strength asymptote (~70–80% at 1 year) tagged #gap/no-fulltext-access.

Supersession check (R25): PubMed searched for senolytic wound-healing RCTs and meta-analyses (2022–2026) — 66 papers found; none qualify as large RCT (n>100) or meta-analysis on this specific senolytic/wound-healing question. Most recent work is preclinical in-vivo (Numani 2026 diabetic mice; ABT-263 dressing 2026) or reviews (Ueda 2026). No supersession candidates. literature-checked-through: 2026-05-19 confirmed.

[2026-05-19] R40 verify — molecules/proteins/col1a1.md

Pages verified: 1

  • molecules/proteins/col1a1.md — flipped to verified: true

Corrections applied (5):

  1. Shuster 1975 PMID wrong: 1220399 → 1220811. PMID 1220399 resolves to a 1975 Russian-language ENT article (Tulebaev & Moshkevich). Correct PMID 1220811 = Shuster S, Black MM, McVitie E, Br J Dermatol 1975;93(6):639–43. Correct DOI 10.1111/j.1365-2133.1975.tb05113.x added.
  2. Varani 2006 n wrong: footnote said “n=~9–10 donors per group” → corrected to n=6 donors per group for in-tissue procollagen (Fig 1), and n=26 isolates from 8 young / 37 isolates from 8 old donors for in-vitro monolayer (Fig 2). These are materially different sample structures.
  3. Varani 2006 “~1%/yr rate” misattribution removed: footnote and body text both implied Varani 2006 “provides experimental basis for the ~1%/yr collagen loss rate.” The paper does NOT quantify or report a per-year rate; it reports a 68% cross-sectional procollagen reduction (82 ± 16 vs 56 ± 8 ng/mm³, old vs young). The ~1%/yr annualized rate derives from Shuster 1975. Corrected in footnote, intro paragraph, and H1 section.
  4. Varani 2006 key figures added to footnote: 68% tissue reduction and in-vitro per-cell values (82 vs 56 ng/5×10⁴ cells) now captured.
  5. Gaps table updated: Shuster 1975 BUG-2 note resolved (correct PMID/DOI now known); Varani 2006 “download pending” resolved (local PDF confirmed 2026-05-19).

Sources checked: UniProt P02452 REST API (canonical length 1464 aa, domain structure confirmed); Fisher 2009 cross-checked via verified study page only (not re-read); Varani 2006 full PDF read (8 pp); Shin 2019 pages 1–4 confirmed collagen percentage figures; Shuster 1975 confirmed via PubMed efetch PMID 1220811.

Unverifiable: Purohit 2016 (doi:10.1016/j.jdermsci.2016.04.004) — closed-access, remains tagged no-fulltext-access.

Supersession check (R25): PubMed search for newer cohort/meta-analysis on dermal collagen decline rate — no large-cohort meta-analysis found superseding the Shuster 1975 / Varani 2006 framing. literature-checked-through: 2026-05-19 confirmed.

Downstream propagation needed:

  • studies/varani-2006-collagen-aged-skin — if/when seeded, ensure n=6 in-tissue / n=26+37 isolates from 8 donors per group and 68% figure are used, not ~9–10; page does not yet exist.
  • Any downstream page that cites Varani 2006 for the “~1%/yr” rate should have that attribution corrected to Shuster 1975.

[2026-05-19] R40 verify — molecules/proteins/mmp-1.md

Pages verified: 1

  • molecules/proteins/mmp-1.md — flipped to verified: true

Corrections applied (6):

  1. Mouse ortholog UniProt accessions wrong: Q9JHH5 → Q9EPL5 (Mmp1a); Q8K157 → Q9EPL6 (Mmp1b). Original accessions resolved to CXCL11_MOUSE and GALM_MOUSE respectively.
  2. Catalytic domain boundary: 100–270 → 100–269 (UniProt P03956 22 kDa chain); hemopexin domain start: 293 → 270 (27 kDa autolytic fragment); four hemopexin repeat positions added (275–324, 325–371, 374–422, 423–466).
  3. Open Targets druggability resolved: removed #gap/needs-canonical-id; confirmed SM Approved Drug: True (max tier 1); aging-context tier 3 retained + documented.
  4. Coppé 2008 SASP claim qualified: full PDF read. MMP-1 NOT in main text or primary figures — appears only in supplemental Table S2 XLS (120-protein catalog). Body and footnote corrected.
  5. MMP-3/MMP-9 paragraphs trimmed: replaced ~150-word duplicate paragraphs with two-sentence cross-links to dedicated pages.
  6. Frontmatter mouse-ortholog updated with confirmed accessions.

Supersession check: PubMed search (MMP-1 + skin aging + meta-analysis/RCT, 2023–2026): 2 results, both on dietary photoprotection botanicals — not superseding Fisher 1996/2009. literature-checked-through: 2026-05-19 confirmed.

Downstream propagation needed: Check [[sasp]] — if it cites MMP-1 as a named Coppé 2008 main-text component, the qualification applies there too.

[2026-05-19] R40 verify — molecules/proteins/timp-1.md

Pages verified: 1

  • molecules/proteins/timp-1.md — flipped to verified: true

Corrections applied (5):

  1. NTR domain boundary wrong: 24–127 → 24–147 (UniProt P01033 feature annotation; C-terminal domain boundary 128–207 → 148–207)
  2. MMP-2 framing corrected: page implied TIMP-1 does not inhibit MMP-2; UniProt confirms TIMP-1 does inhibit MMP-2 in isolation; nuanced to “TIMP-2 is the dominant regulator of MMP-2 activation via the MT1-MMP ternary complex; TIMP-1 does not inhibit MT1-MMP (MMP-14)”
  3. MMP-12 and MMP-16 missing from target list: added per UniProt curated experimental evidence
  4. Nagase 2006 DOI wrong: 10.1093/cvr/cvl248 (invalid — 404) → 10.1016/j.cardiores.2005.12.002 (Cardiovasc Res 2006;69:562–573)
  5. MT-MMP resistance clarification: updated Key gap section — resistance is specific to MMP-14 (MT1-MMP), not all MT-MMPs; MMP-16 (MT3-MMP) is a TIMP-1 substrate

Sources: Fisher-1996 + Fisher-2009 primary claims cross-checked against verified study pages (both R39-verified). UniProt P01033 canonical fields verified via REST API.

Supersession check: No meta-analyses or large RCTs on TIMP-1 in aging connective tissue post-2023. literature-checked-through: 2026-05-19 already set on page.

Downstream propagation needed: None identified — no other wiki pages contain domain boundary numbers for TIMP-1 specifically. The MMP-2 nuance and Nagase DOI are self-contained to this page.

[2026-05-09] ingest | engineered-negligible-senescence batch (5 pages)

Spawned-from: user question turn 2 — what would a holistic anti-aging solution look like? interventions that fully arrest hallmark accumulation. The engineered-negligible-senescence (ENS) framing identified five seeding candidates which were dispatched in this batch.

Added (5 pages):

  1. frameworks/engineered-negligible-senescence.md — synthesis MOC for the ENS framing. Aggregates per-hallmark arrest requirements, three strategic paradigms (SENS damage repair / hyperfunction program suppression / comparative-biology mechanism transfer), the cancer-problem callout. Framework page; no verified discipline. Cross-links 13 hallmark pages, hallmark-causality-graph, sens-damage-categories, all five relevant hypothesis pages, three model-organism pages.

  2. frameworks/cancer-aging-tradeoffs.md — synthesis MOC for the cancer-longevity trade-off pattern. Six-row trade-off table; comparative biology of NMR/elephant/bowhead cell-autonomous tumor suppression; validated mouse template (Tomás-Loba 2008 Cell, doi:10.1016/j.cell.2008.09.034) + García-Cao 2002 EMBO J + Matheu 2004 Genes Dev. All three primary-source DOIs PubMed-verified. Framework page.

  3. hypotheses/longevity-escape-velocity.md — verified: true. Mode B (conceptual-frame) hypothesis. status: contested. Recency search 2020–2026 ran via PubMed; 6 results, none RCT or meta-analysis. Verifier corrections: Hayflick mis-attributed as Warner 2005 EMBO Reports signatory (corrected to Olshansky/Hayflick/Carnes 2002 J Gerontol — Hayflick is on Olshansky 2002, not Warner 2005); de Grey 2004 PLoS Biology AEV threshold mis-stated as “+1 year of life expectancy per year” (corrected to ~10 percent per year sustained decline in age-specific mortality rates); Palmer 2022 three-tier framing mis-stated and corrected. Verified-scope: de Grey 2004 + Warner 2005 + Palmer 2022 against PDFs; Oeppen-Vaupel 2002 + Olshansky 2002 + Lewis-de Grey 2024 abstract-level only (not_oa).

  4. processes/lipofuscin.md — verified: false (permanent blocker). 7 cited primary sources: 5 not_oa, 2 download-failed (Brunk & Terman 2002 Eur J Biochem + Wu & Sparrow 2010 JACS — failed both). Bibliographic metadata cross-checked via Crossref for all 7. Verifier corrections: Moore 1995 caloric-restriction lipofuscin-retarding dose was 52 percent CR (not 25 percent mild CR); supersession candidate flagged — Chiavacci 2026 Aging Cell (PMID 42024652) reports massive cardiomyocyte lipofuscin in 300-yr Greenland shark coexisting with preserved cardiac function, challenging the canonical lipofuscin → cell death framing. ⚠️ banner stays; verified-scope: null. Lint pass should treat this page as a permanent-blocker #gap/no-fulltext-access rather than a follow-up task.

  5. processes/glucosepane.md — verified: true (abstract-level scope). 14 cited primary sources, all DOIs Crossref-verified. Verifier corrections (4): Sell 2005 tissue scope was skin (n=110) + glomerular basement membrane (n=28) — seeder had fabricated “tendon” (corrected); Sell 2005 quantitative ratios qualified as review-synthesis figure (~10–100× pentosidine ratio not directly stated in Sell 2005 abstract; expanded with actual measurements ~2000 pmol/mg in nondiabetic age 90, ~5000 pmol/mg in diabetic, p<0.0001); Monnier 2022 measured plasma protein-bound AGEs (not skin) — CSME association is CML-driven, not glucosepane-specific; Hartog 2011 BENEFICIAL Phase 3 details enriched (peak VO2 P=0.06; diastolic P=0.32; first-author initials JW not JWL). Recency search 2020–2026 yielded 17 papers; no supersession.

Gaps surfaced:

  • Two-time recurrence of seeder-fabricates-outcomes pattern (R29 feedback memory): Sell 2005 tissue-scope fabrication caught by glucosepane verifier. The seeder gets paper target/mechanism right but invents specific tissue scope or experimental endpoints. Verifier briefs should continue to emphasize “verify outcomes-not-just-targets” + tissue-scope cross-checks.
  • Lipofuscin verification permanently blocked. The two foundational Brunk & Terman 2002 reviews are not_oa with download-failure; centrophenoxine (Terman 1999) is also not_oa. The page is correctly flagged verified: false with #gap/no-fulltext-access.
  • Greenland shark cardiomyocyte lipofuscin (Chiavacci 2026) is a cross-cutting supersession point; already cited in negligible-senescence (verified); now also flagged on lipofuscin.

Propagation pass not yet run. Inbound-link counts for the new framework pages will need to be updated as downstream hallmark and intervention pages start cross-linking to engineered-negligible-senescence and cancer-aging-tradeoffs in future batches.

[2026-05-09] ingest | SENS-research current-state batch (8 pages + MOC update)

Spawned-from: user request — Review the latest SENS research and make sure the wiki is up to date with the latest understanding related to the various needed interventions. Document current research in progress per framework item. The SENS damage-categories MOC at frameworks/sens-damage-categories.md was thin (just a 7-row maturity table) and dated; user asked for a comprehensive 2024-2026 research-in-progress update across all seven categories.

Updated (1 framework MOC):

  • frameworks/sens-damage-categories.md — substantially expanded. Added: (1) section on SENS Foundation/de Grey 2021 organizational rupture and LEV Foundation 2022 founding; (2) maturity matrix table (Phase 3/Phase 1/preclinical/conceptual per category); (3) per-category current-research-in-progress sections for all 7 categories with verified PMIDs/DOIs and clinical-trial NCT numbers where known. ~310 lines added, including: RepleniSENS (bemdaneprocel Phase 3 first-patient Sept 2025; Sumitomo iPSC-DA Japanese MAA Aug 2025; Vertex zimislecel pivotal; LyGenesis Phase 2a; TRIIM-X enrolling); OncoSENS (imetelstat FDA 2024 for MDS — adjacent precedent; WILT remains conceptual; cancer-resistance retrogene transfer preclinical); ApoptoSENS (UBX1325 BEHOLD positive NEJM Evidence 2025 / ASPIRE Phase 2b miss May 2025; Rubedo RLS-1496 Phase 1; Oisin FAST-PLV $15M Series A; DT2216 Phase 1); MitoSENS (Lewis 2024 first-in-vivo ATP8 allotopic in mouse — central milestone); LysoSENS (Cyclarity UDP-003 Phase 1 dosing Jan/Feb 2025 in Australia — first clinical LysoSENS); AmyloSENS (lecanemab/donanemab/acoramidis/tafamidis/patisiran/vutrisiran approvals; prasinezumab Phase 3 advancing despite PADOVA primary miss); GlycoSENS (Revel Pharmaceuticals preclinical glucosepane breakers; alagebrium failed; no clinical-stage program).

Added (8 pages):

  1. molecules/compounds/lecanemab.md — verified: true (partial scope: Clarity AD primary van Dyck 2023 NEJM PDF verified end-to-end; OLE + safety meta abstract-only). FDA-approved AmyloSENS flagship; Aβ protofibril mAb. Verifier corrections (7): ARIA-H placebo rate 2.7→9.0% (seeder mis-assigned ARIA-E placebo rate to ARIA-H row); ARIA-H lecanemab rate 13.0→17.3%; falls rates clarified for both arms; symptomatic ARIA-E added (2.8% lecanemab vs 0% placebo); APOE4-stratified ARIA-E rates added (5.4%/10.9%/32.6% noncarrier/heterozygote/homozygote); brain volume claim corrected (van Dyck 2023 explicitly states volumetric MRI not fully analyzed in primary paper); QoL secondary % removed (not in source). Schema: translation-gap: effect-size-modest → human-evidence-strong (effect-size-modest not in canonical CLAUDE.md enum).

  2. molecules/compounds/donanemab.md — verified: true (partial scope: Sims 2023 JAMA primary + Salloway 2025 Alz Dement TBZ-ALZ4 PDFs verified end-to-end; Zimmer 2025/2026 + Lu 2025 + Yaari 2025 abstract-only). FDA-approved AmyloSENS Aβ-pyroglutamate mAb. Verifier corrections (7): ARIA rates precision (5.8→6.1; 31.3→31.4; placebo 13→13.6%); amyloid clearance source corrected (Sims 2023 not Zimmer 2026); Cochrane 2026 antibody count 9→7 per abstract; iADRS LSM precision; ARIA discrepancy between Sims (24.0%) and Zimmer combined-trials (24.4%) clarified; same translation-gap schema fix as lecanemab.

  3. molecules/compounds/acoramidis.md — verified: true (partial scope: Judge 2025 JACC + Judge 2025 Circulation OLE PDFs verified end-to-end; Gillmore 2024 NEJM ATTRibute-CM primary abstract-only — failed). Most recent (Nov 2024) AmyloSENS TTR-stabilizer FDA approval. Verifier corrections (5): Win ratio CI added (95% CI 1.4–2.2); 35.9%/50.5% time-to-event proportions correctly attributed to Judge 2025 JACC secondary analysis (not Gillmore 2024 primary win-ratio paper); annualized CVH frequency CI added (95% CI 36–64%); two footnotes upgraded from abstract to full-PDF-verified.

  4. molecules/compounds/tafamidis.md — verified: true (partial scope: ATTR-ACT Maurer 2018 NEJM + Coelho 2012 Neurology + Facin 2025 meta-analysis PDFs verified end-to-end; Damy 2025 OLE + Judge 2025 acoramidis OLE + Fontana 2025 HELIOS-B + Hussain 2025 + Bulawa 2012 abstract-only or not_oa). FDA-approved 2019 ATTR-CM TTR-stabilizer; foundational AmyloSENS precedent. Verifier corrections (6): Coelho 2012 randomized n=128 (65/63), ITT n=125 — seeder had 64/64; ITT NIS-LL responder rates added (45.3% vs 29.5% p=0.068); ATTR-ACT primary win ratio 1.695 (95% CI 1.255–2.289) added; CV hospitalization precision (RRR 0.68 95% CI 0.56–0.81); Facin 2025 attribution clarified — RR 0.71/0.81 are TTR-stabilizer subgroup pooled (tafamidis + acoramidis) not tafamidis alone.

  5. molecules/compounds/patisiran.md — verified: true (partial scope: APOLLO + APOLLO-B + 5-yr OLE Adams 2025 + Aimo 2025 NMA PDFs verified end-to-end; HELIOS-B Fontana 2025 abstract-only — local PDF was misfetched DI-fusion help guide; archive bug logged at ~/dev/a local paper archive/FEATURE_REQUESTS.md BUG-7). FDA-approved 2018 first-ever RNAi therapeutic. Most substantive verifier correction batch this session — 14 corrections, including multiple seeder-fabricated quantitative claims:

    • Infusion duration 60→80 min
    • TTR knockdown precision: ~80% → median 81% range −38 to 95%
    • Peripheral edema 29→30%
    • Trial geography 49 countries → 44 sites/19 countries (fabricated number)
    • APOLLO-B randomization 2:1→1:1 (fabricated by seeder; APOLLO-B was 1:1)
    • APOLLO-B WT% 56→80% (fabricated)
    • APOLLO-B 6MWT primary endpoint figures completely wrong — corrected from “+13.0 m, 95% CI 0.5–25.4, p=0.040” to actual H-L estimate +14.69 m (95% CI 0.69–28.69; P=0.02)
    • HELIOS-B randomization 2:1→1:1
    • HELIOS-B primary endpoint RR 0.72 → HR 0.72 (statistic-type error)
    • HELIOS-B all-cause mortality RR 0.64 → HR 0.65 through 42 months
    • HELIOS-B 6MWT LS mean +24.7→+26.5 m
    • 5-yr OLE enrollment corrected (211/212 eligible → 138 completed → 41 (19.4%) died)
    • Aimo 2025 NMA framing corrected — acoramidis HR 0.76 (95% CrI 0.56–1.03) was NOT significant; only tafamidis and vutrisiran monotherapy reached significance vs placebo

  6. processes/7-ketocholesterol.md — verified: false (no verifier dispatched this batch; deferred). LysoSENS substrate; Cyclarity UDP-003 Phase 1 lead. Seeder integrated 9 PMIDs; UDP-003 trial start date corrected by seeder to February 25, 2025 per ClinicalTrials.gov NCT06813339 (brief had January 2025 from press releases).

  7. processes/allotopic-expression.md — verified: false (no verifier dispatched this batch; deferred). MitoSENS canonical mechanism. Lewis 2024 in vivo ATP8 mouse milestone (PMID 39659757) integrated as headline. Seeder also integrated yeast 1990s, Manfredi 2002 Nat Genet, Oca-Cossio 2003 Genetics, Boominathan 2016 NAR ATP6+ATP8 co-expression, Lewis 2020 Redox Biol codon optimization.

  8. processes/advanced-glycation-end-products.md — verified: false (no verifier dispatched this batch; deferred). GlycoSENS parent class for glucosepane. 14 PMIDs integrated; Granic 2023 sarcopenia systematic review + Flensted-Jensen 2026 RCT muscle AGE distribution as recent recency-search additions.

Cross-cutting findings integrated into MOC:

  • Cochrane 2026 anti-amyloid meta-analysis (Nonino et al., PMID 41985900, doi:10.1002/14651858.CD016297; 17 RCTs / 7 antibodies / n=20,342) concluded effects on cognition are “trivial” and effects on dementia severity are “little to no meaningful difference.” Substantive supersession finding integrated into both lecanemab and donanemab (#gap/contradictory-evidence tags) and into the SENS MOC AmyloSENS section as a dedicated discordance callout. The 2050 ceiling for loss-of-proteostasis arrest may be lower than originally estimated unless next-generation tau-targeting (E2814, BIIB080) delivers substantively larger effects.

  • Schema gap surfaced (recurring): Both lecanemab and donanemab seeders defaulted to translation-gap: effect-size-modest which is not in the canonical CLAUDE.md enum (phase-3-rct-needed | biomarker-only | preclinical-only | human-evidence-strong | translation-blocked-safety | translation-blocked-cost). Both verifiers defaulted to human-evidence-strong with inline notes. Recommendation for CLAUDE.md update: consider adding effect-size-modest or clinically-contested as a canonical enum value for FDA-approved drugs whose primary controversy is MCID rather than regulatory status.

  • Seeder-fabricates-outcomes pattern (R29) at higher severity than prior instances. The patisiran case is the most substantial fabrication caught to date — APOLLO-B randomization (1:1 vs fabricated 2:1), WT% (80% vs fabricated 56%), HELIOS-B statistic-types (HR vs fabricated RR), trial geography (44 sites/19 countries vs fabricated 49 countries). The seeder gets the paper target and overall mechanism correct but invents specific quantitative endpoints. Recommendation: verifier briefs should continue to emphasize “verify outcomes-not-just-targets,” with specific attention to randomization ratios, primary endpoint statistic-type (HR/RR/OR), and sample-size figures — these are the most-fabricated fields in this session’s data.

  • Five compound-page verifications in single batch is at the upper end of feasibility. Verifiers ran ~3.5–10 minutes each; the patisiran verifier took 10+ minutes due to PDF re-fetching. Future batches at this scale should dispatch in two waves (3+2 or 4+4) to reduce parallel-execution risk if any single verifier hangs.

Deferred (3 process pages without verifier this session): 7-ketocholesterol, allotopic-expression, advanced-glycation-end-products — verified: false carried forward; primary-source PDFs largely closed-access or pending download; these pages have lower load-bearing quantitative claims than the FDA-approved compound pages.

Propagation pass not run. Future batches should add inbound-link counts for the 5 newly seeded compound pages and re-check downstream hallmark/intervention pages for cross-link integrity.

[2026-05-12] ingest | Zhang et al. 2023 Nature — naked-mole-rat HAS2 transgenic mice (xenogeneic gene transfer)

User-requested ad-hoc ingestion of doi:10.1038/s41586-023-06463-0 (PMID 37612507; PMC10666664). The paper is the first whole-animal demonstration that an NMR longevity adaptation can be exported to another mammalian species — establishing xenogeneic / interspecies gene transfer of longevity mechanisms as a defined intervention modality. Reframings beyond the obvious NMR-protein page anticipated and surfaced before dispatch.

New atomic pages (4; all verified: true after wiki-verifier pass against PMC HTML + local PDFs for cross-references):

  • studies/zhang-2023-nmrhas2-mouse-healthspan.md — primary study; 4 substantive corrections during verification (cancer >27-mo subgroup n’s, tamoxifen route, DMBA/TPA per-sex n’s, del Marmol 2021 footnote rewrite)
  • molecules/proteins/has2.md — hyaluronan synthase 2; UniProt Q92819 + 552 aa + 7-TM + Mg²⁺ + PTM sites; Has2-null embryonic lethality stage corrected to E9.5–10 per Camenisch 2000; Asn→Ser substitution positions remain #gap/needs-verification (not in Tian 2013 main text)
  • molecules/compounds/hyaluronic-acid.md — HA biology; MW-direction-of-effect (HMW = anti-inflammatory; LMW = pro-inflammatory via TLR4); polymer with no canonical PubChem CID (#gap/needs-canonical-id documented); R16 framework added two new mechanism classes (extracellular-matrix-remodeling, immunomodulation) to intervention-classes.md
  • molecules/proteins/cd44.md — primary HA receptor; Tian 2013 mechanism corrected from “CD44 siRNA” to “CD44 antibody-blockade”; Takasugi 2023 ATF6 mechanism corrected to HA-independent

Propagation pass (10 existing pages):

  • model-organisms/heterocephalus-glaber.md — added Zhang 2023 to HMW-HA section + new “Transferability of NMR mechanisms to other species” section + cross-references; reframed Tian 2013 species-specific narrative (mouse Has2 alone reproduces benefits → HA abundance, not Asn→Ser substitutions, is the operative variable)
  • hallmarks/chronic-inflammation.md — added “HMW-HA augmentation — comparative-biology-derived intervention class” subsection in Therapeutic landscape; added Termeer 2002 LMW-HA → TLR4 mechanism citation
  • processes/gut-barrier.md — added HMW-HA / HAS2 row to interventions table + dedicated “HMW-HA and gut barrier preservation — Zhang 2023 evidence” subsection (FITC-dextran, goblet/Paneth counts, organoid HMW-HA vs LMW-HA rescue at 20 µg/mL)
  • frameworks/cancer-aging-tradeoffs.md — expanded existing Zhang 2023 reference with full quantitative detail and the “mouse Has2 alone” reframing; added study-page wikilink + footnote
  • microbiome/firmicutes-bacteroidetes-ratio.md — added Zhang 2023 microbiome shift (B/F ratio direction in old nmrHas2 mice consistent with Wilmanski 2021); explicitly caveated as not rehabilitating phylum-level F/B as a useful biomarker
  • biomarkers/frailty-index.md — added Zhang 2023 Whitehead 31-parameter FI evidence as preclinical intervention-responsive signal alongside ITP rapamycin/acarbose
  • hypotheses/negligible-senescence.md — added Zhang 2023 to 2025–2026 evidence updates; updated “What would update this hypothesis” transgenic-transfer criterion noting partial fulfillment (transferability confirmed; non-Gompertzian mortality NOT extended to mice)
  • frameworks/sens-damage-categories.md — expanded existing Zhang reference with study-page wikilink + full numerics
  • phenotypes/cancer.md — added “Comparative biology — exporting cancer-resistance mechanisms across species” section integrating Zhang 2023 + cross-references to NMR/HAS2/HA/CD44/cancer-aging-tradeoffs
  • frameworks/engineered-negligible-senescence.md — already cited Zhang 2023 (line 38); no edit required

Gaps surfaced:

  • #gap/needs-verification retained for NMR HAS2 Asn→Ser substitution positions (Tian 2013 main text doesn’t state; supplementary sequence alignment not locally available)
  • #gap/needs-canonical-id for hyaluronic-acid PubChem CID (polymer; no canonical small-molecule CID)
  • #gap/needs-human-replication for AAV-HAS2 / HMW-HA-augmentation translational pathway in humans (preclinical only as of 2026)
  • #gap/contradictory-evidence retained for Tian 2013 vs del Marmol 2021 ultra-HMW MW dispute (Zhang 2023 partially relaxes the load-bearing role of the absolute MW figure but does not resolve it)

Local-archive feature gap surfaced: PMC author-manuscript deposits not in PMC OA-package distribution currently fail (no candidate URLs after filtering). Verification proceeded via WebFetch against PMC HTML. Worth noting in a local paper archive/FEATURE_REQUESTS.md at next review.

Conceptual reframings persisted to memory (feedback_paper_impact_broad.md): when ingesting a new paper, enumerate the conceptual frames it reframes (intervention modality, hypothesis, comparative-biology) not just the obvious subject-page. The user surfaced this correction early — original framing was “naked-mole-rat content” only; broader reframing as xenogeneic-gene-transfer modality was the productive one.

[2026-05-15] ingest | Trętowicz 2026 Nat Metab — whole-blood NAD+ stable with age

User shared URL to new Nature Metabolism paper (published 2026-05-14, day-old). Resolved a typo in the URL fragment (-6 → -5), confirmed paper exists, retrieved Abstract + 4-page Reporting Summary PDF + 8 source-data XLSX files via Springer ESM endpoint. Main-text Results/Discussion paywalled (Unpaywall: oa_status: closed); a local paper archive cannot queue download (DOI not yet in OpenAlex snapshot).

Paper: Trętowicz MM, Scantlebery AML, Schomakers BV, et al. Human whole-blood NAD+ levels do not vary with age or lifestyle interventions. Nat Metab (2026). DOI: 10.1038/s42255-026-01537-5. Senior author Houtkooper (Amsterdam UMC); peer-reviewed by Covarrubias, Migaud, Wu. EU NADIS consortium funded; no competing interests.

Headline finding: Whole-blood NAD+ is stable with age across 7 independent cohorts (n=303 total). All six independent age comparisons null: P=0.24 dedicated aging cohort (n=40); R²=0.012 CardioHT (n=26); P=0.50 ELITE older-vs-athlete (n=47); R²=0.051 LLS (n=70); P=0.62 TEAMS exercise+protein (n=65); P=0.62 MEJNES2019 frail older adults (n=31). Same UHPLC-HRMS assay readily detects NR-supplementation-induced NAD+ elevation in twin-pair positive-control cohort (n=24, NCT03951285) — so the null is not a sensitivity artifact. Prior conflicting blood-NAD+/aging literature attributed to pre-analytical handling artifacts (frozen-without-methanol samples lose 30–80% of NAD+).

Pages added:

  • studies/tretowicz-2026-blood-nad-stable-aging.md — new study page; verified: false (Results/Discussion paywalled; scope notes record source-data + reporting-summary verification of methods, n’s, cohort registries, per-cohort P/R²)
  • biomarkers/nad-blood-biomarker.md — new biomarker page documenting whole-blood NAD+ as a refuted aging biomarker; human-evidence-level: limited-negative; covers the original biomarker hypothesis, the prior positive literature (Chaleckis 2016, Yang 2022, Wang 2023, Breton 2020), the Trętowicz 2026 refutation with per-cohort statistics, sources of confusion (PBMC vs whole-blood vs plasma; freeze-thaw artifacts; NAD+/NADH interconversion; NADP+ confounding), and what the negative result does NOT rule out (tissue-level NAD+ decline, flux/turnover, disease-specific deficits, pharmacodynamic-biomarker validity for NR engagement); verified: false

Pages updated (propagation pass):

  • interventions/pharmacological/nad-precursors.md — rewrote ”## NAD+ Declines with Age” section to distinguish blood (refuted by Trętowicz 2026) from tissue (Janssens 2022 muscle + Massudi 2012 skin — still valid); added Trętowicz to “2024–2026 RCT and meta-analysis wave” as biomarker-validity entry; updated class-level interpretation; literature-checked-through: 2026-05-08 → 2026-05-15; verified-scope extended (R37)
  • molecules/compounds/nmn.md — added blood-vs-tissue distinction paragraph under “Why NAD+ declines with age” with #gap/contradictory-evidence flag; new footnote [^tretowicz2026]; literature-checked-through: 2026-05-15
  • molecules/compounds/nr.md — added “rationale-undermined-at-blood-level” bullet to Limitations and Gaps; rewrote tissue-specific-NAD-dynamics bullet to note blood NAD+ is no longer a useful aging biomarker; new footnote [^tretowicz2026]; literature-checked-through: 2026-05-15

Gaps surfaced:

  • #gap/contradictory-evidence — blood NAD+ (stable with age, Trętowicz 2026) vs tissue NAD+ (declines in skin + muscle, Massudi 2012 + Janssens 2022). Functional implication is open: is the relevant aging-NAD+ phenomenon tissue-specific intracellular pool depletion, or NAD+ flux/turnover at constant pool size?
  • #gap/needs-tissue-level-replication — re-run Trętowicz’s validated UHPLC-HRMS method on muscle/liver/brain biopsy banks to test whether the tissue-decline reports survive the same pre-analytical discipline
  • #gap/no-fulltext-access — Results + Discussion text paywalled; verifier flip blocked until either (a) a local paper archive ingests the DOI when OpenAlex propagates, or (b) author share link / institutional access becomes available

Wiki structure delta: A new biomarkers/nad-blood.md page is a natural follow-up (R37 candidate) — the wiki has 19 biomarker pages, and “blood NAD+” is now a relevant negative biomarker. Deferred pending user direction (would re-cite Trętowicz 2026 as primary evidence for the null aging-biomarker claim).

Process notes:

  • User’s URL had a one-digit typo (-6 → -5); my first WebFetch returned a 303-to-SSO that looked like paywall-protected content but actually masked a 404. Lesson: when a Nature URL appears to require auth, fetch with a browser UA + cookie file to disambiguate “paywall” from “doesn’t exist” — the user-agent-sensitive 404 was diagnostic.
  • DOI registration lag is real: paper published 2026-05-14, DOI not in Crossref/PubMed/Europe PMC/OpenAlex as of 2026-05-15. The Springer ESM endpoint (static-content.springer.com/esm/art%3A.../MediaObjects/...) serves supplementary materials publicly even when the main article is paywalled — confirmed for this paper. Reporting Summary PDF (~3.3 MB) + per-figure Source Data XLSX files are the single highest-leverage way to verify methods + numerics without main-text access.
  • Source-data XLSX files include author-computed P-values and R² in the bottom row of each sheet — these are direct primary-source numerics, not seeder memory.

[2026-05-19] verify | molecules/proteins/mmp-3.md (R40)

Pages verified: 1 — molecules/proteins/mmp-3.md. Flipped to verified: true.

Sources checked:

  • Fisher 1996 (doi:10.1038/379335a0) — cross-checked against verified studies/fisher-1996-photoaging-ap1-mmp.md; PDF not re-read.
  • Fisher 2009 (doi:10.2353/ajpath.2009.080599) — cross-checked against verified studies/fisher-2009-collagen-fragmentation-mmp.md; PDF not re-read.
  • UniProt P08254 — verified via REST API 2026-05-19.
  • Sumsuzzman 2022 — PMID 36087701 confirmed via PubMed efetch.
  • Mak 2025 — PMID 40998199 + DOI confirmed via Crossref + PubMed efetch.

Corrections applied (5):

  1. Sumsuzzman 2022 DOI: 10.1016/j.arr.2022.101730 → 10.1016/j.arr.2022.101729 (article number off by 1; confirmed via PMID 36087701 efetch)
  2. Mak 2025 DOI: incomplete doi:10.1016/j.brainres.2025 → full doi:10.1016/j.brainres.2025.149954; added confirmed n=1316 (12 studies, 806 AD, 510 controls); removed #gap/needs-replication — full DOI not confirmed tag
  3. Catalytic domain boundary: 100–266 → 100–272 per UniProt P08254 feature table
  4. MMP-3 sub-erythemogenic dose-threshold claim corrected: original text implied 0.01 MED threshold applied to MMP-3; Fisher 1996 Fig. 2 dose-response is for MMP-1 and MMP-9 only; MMP-3 mRNA dose-response not independently characterised — added gap tag + clarification
  5. Neurodegeneration section: added confirmed n=1316 (replacing vague “n studies = multiple cohorts”)

Supersession check (R25): PubMed 2022–2026 MMP3+aging+meta-analysis search: 2 results (Sumsuzzman 2022 already cited; telomere-MMP-3 MR paper unrelated). No supersession candidates. literature-checked-through: 2026-05-19 confirmed.

Mouse ortholog Mmp3: confirmed via UniProt P28862 (477 aa; Mus musculus stromelysin-1).

[2026-05-19] verify | molecules/proteins/kitlg.md (R40)

Pages verified: 1 — molecules/proteins/kitlg.md. Flipped to verified: true.

Sources checked:

  • Hattori 2004 (doi:10.1111/j.0022-202x.2004.22503.x) — cross-checked against verified studies/hattori-2004-scf-solar-lentigo.md; PDF not re-read (study page verified: true, 2026-05-19, full PDF read).
  • Young 2024 (doi:10.1182/blood.2024024275, PMC11302459) — full PDF read (14 pp); paper freshly downloaded from PMC.
  • Imokawa 2019 (doi:10.3390/ijms20153666) — PDF read (pp 1–8 of 16; all SCF/KIT claims on this page covered).
  • UniProt P21583 — verified via REST API 2026-05-19.
  • NCBI Gene 4254 — chromosomal location 12q21.32 confirmed.

Corrections applied (6):

  1. Spelling: pleiotrophic → pleiotropic (opening paragraph)
  2. Isoform table: added explicit aa-length column — SCF248 = 248 aa post-signal-peptide (273 aa precursor, signal peptide aa 1–25, TM aa 215–237); SCF220 = 220 aa post-signal-peptide (245 aa precursor, exon-6 deletion replaces aa 174–202 with single Gly removing 28 aa)
  3. Young 2024 footnote: n=multiple mouse cohorts → n=5 young (2 mo) + 9 middle-aged (14 mo) C57BL/6J female mice (scRNA-seq main cohort); expanded mechanism description to include differentiation-inactive HSC signature and reduced lymphoid priming
  4. Imokawa 2019 discrepancy note added: review states c-KIT mRNA 2.14-fold (citing Hattori 2004) — contradicts the verified primary source (1.16-fold NS); canonical 1.16-fold NS value retained on this page
  5. [[hair-greying]] cross-reference: removed erroneous #gap/needs-canonical-id (page not yet seeded) — file confirmed at phenotypes/hair-greying.md
  6. [[kit]] cross-reference: removed erroneous #gap/needs-canonical-id (page not yet seeded) — file confirmed at molecules/proteins/kit.md

Supersession check (R25): PubMed 2022–2026 searches: (a) KITLG+aging+hematopoietic: 0 results; (b) solar lentigo+SCF+keratinocyte+melanocyte: 1 result (PMID 39780507, Gao & Xiang 2025 review on fibroblast roles — not a supersession of Hattori 2004). No supersession candidates. literature-checked-through: 2026-05-19 confirmed in frontmatter.

Unverifiable claims: Nishimura 2011 (doi:10.1111/j.1755-148x.2011.00855.x) — not OA; McSC/hair-greying claims sourced to this review remain unverified against primary source. No quantitative claims sourced exclusively from Nishimura 2011.

Downstream propagation: No corrections required to study pages. [[kit]] and [[hair-greying]] pages may benefit from reciprocal KITLG cross-links (seeding task, not correction propagation).

Graph View

Backlinks