log/R35.md — Round 35 entries

Sub-file of log — see parent for index.

[2026-05-09] verify — molecules/proteins/growth-hormone-receptor.md

Pages verified: 1

molecules/proteins/growth-hormone-receptor.md — corrections: 10 substantive; verified: true (partial scope — Guevara-Aguirre 2011 via PMC full text; Aguiar-Oliveira 2019 via local PDF; Bartke 2003 + Coschigano 2003 abstract only via PubMed efetch; Brown-Borg 1996 confirmed via PMID, no abstract; Leung 1987 confirmed via PMID; Laron 2008 archive mapping error; canonical IDs spot-checked)

Sources checked:

Guevara-Aguirre 2011 (doi:10.1126/scitranslmed.3001845) — PMC3357623 full text fetched; verified
Aguiar-Oliveira & Bartke 2019 (doi:10.1210/er.2018-00216) — downloaded PMC6416709; pp 1–22 read
Coschigano 2003 (doi:10.1210/en.2003-0374) — not_oa; abstract via PubMed PMID 12933651
Bartke 2003 (doi:10.1159/000073704) — not_oa; abstract via PubMed PMID 14583653
Brown-Borg 1996 (doi:10.1038/384033a0) — not_oa; PMID 8900272 confirmed; no abstract (1-page letter)
Leung 1987 (doi:10.1038/330537a0) — not_oa; PMID 2825030 confirmed via PubMed
Laron 2008 (doi:10.14310/horm.2002.1196) — wrong paper (Psyrogiannis et al. iron overload DM2); PMID 18359741 not verified
NCBI Gene 2690 — confirmed GHR, Homo sapiens, HGNC:4263
UniProt P10912 — not independently re-queried (archive mapping error blocked laron.2008; other IDs confirmed via NCBI)

Corrections made (10 substantive):

Cancer claim for Guevara-Aguirre 2011: “~5-fold lower cancer incidence” → actual paper reports 1 cancer case in 99 GHRD subjects vs 17% prevalence in relatives; cancer = 0% of GHRD deaths vs 20% of relative deaths (p=0.003); the “5-fold” figure is not stated in the paper
Diabetes claim: “~5% in relatives” → paper states 5% of deaths and 6% of all diseases in relatives; 0/90 living GHRD subjects (95% CI 0–4%; p=0.02)
Cohort n: “n≈100 affected” → n=99 GHRD subjects monitored since 1988 vs 1,606 first-to-fourth-degree relatives; serum subgroup n=16 GHRD vs 13 relatives
Added exact IGF-1 and HOMA-IR figures from Guevara-Aguirre 2011: IGF-I ≤20 ng/mL (GHRD) vs mean 144 ng/mL (relatives); HOMA-IR 0.34 vs 0.96 (p<0.05)
Added mortality breakdown: 21 of 30 deaths in GHRD subjects over age 10 were non-age-related (not just “accidental death confounds”)
Pre-formed dimer citation removed from Leung 1987: Leung 1987 is a cloning paper and does not characterize dimerization state; tagged needs-citation pointing to de Vos 1992 / Brown 2005
JAK2-STAT5 citation removed from Leung 1987: cloning paper does not characterize downstream signaling; tagged needs-citation
GHR-KO lifespan “~35–40%”: removed unverified percentage; Coschigano 2003 abstract confirms significant extension but gives no quantitative percentage
GH-transgenic “50–60% reduction”: removed unverified percentage; Bartke 2003 abstract says “drastically shortened” + kidney pathology + cognitive aging; specific % not in abstract
Ames dwarf “30–50% longer”: removed unverified percentage; Brown-Borg 1996 is a 1-page letter with no available abstract
Somatopause “~14% per decade”: removed as unsourced; tagged needs-citation
Liver IGF-1 “~75%”: removed as unsourced; tagged needs-citation
Laron 2008 archive mapping error flagged in footnote
Aguiar-Oliveira 2019 footnote: removed “pending download” (downloaded); added actual content (IGHD Brazilian cohort n=105, longevity data, Laron comparisons)
Leung 1987 footnote: corrected “rabbit liver and human” to “rabbit liver; rabbit serum; human cDNA clones”; added note that paper covers cloning only

Archive mapping bug:

doi:10.14310/horm.2002.1196 (Laron Z, Hormones 2008) — Psyrogiannis et al. 2003 (Hormones 2(3):161-168) instead. This is a DOI→URL mapping error in a local paper archive. Recommend filing FEATURE_REQUESTS.

Downstream propagation needed:

No other pages found citing growth-hormone-receptor.md study page directly; entity pages igf-1, igf-1-biomarker, insulin-igf1, laron-syndrome may contain GHR-derived claims that should be cross-checked against the corrected figures here.

[2026-05-09] verify — molecules/proteins/c-met.md

Pages verified: 1

molecules/proteins/c-met.md — corrections: 8 substantive; verified: true (partial scope — Birchmeier 2003, Koike 2025, Park 1987, Paik 2020 read; Tatsumi 2010 + Bottaro 1991 not_oa)

Sources checked:

Birchmeier 2003 (doi:10.1038/nrm1261) — local; full 11 pp read
Koike 2025 (doi:10.1111/acel.70042) — local; full 17 pp read
Park 1987 (doi:10.1073/pnas.84.18.6379) — downloaded; full 5 pp read
Paik 2020 (doi:10.1056/NEJMoa2004407) — downloaded; full 13 pp read
UniProt P08581 — phosphosites, chains, domains, Ensembl confirmed via REST API
Tatsumi 2010 (doi:10.1111/j.1740-0929.2009.00712.x) — not_oa; not verified
Bottaro 1991 (doi:10.1126/science.1846706) — not_oa; not verified

Corrections made (8 substantive):

ORR figure for tepotinib (Paik 2020): 43% → “46% (95% CI 36–57) by independent review; 68% in treatment-naive patients.” The 43% figure does not appear in the paper.
Trial design for Paik 2020: “rct-phase-2” → “phase-2, single-arm, open-label (VISION trial).” Not randomized.
Footnote for Paik 2020: added n=99 efficacy population, 95% CIs, median duration of response 11.1 mo, and removed “808 citations; pending download.”
Ser985 juxtamembrane inhibitory site removed: not present in UniProt P08581 or Birchmeier 2003. Replaced with correct Tyr1003 (Cbl docking site, eliminated by METex14) with primary-source support.
Gab1 recruitment mechanism corrected: “via Grb2-SH2 interactions” → Gab1 has its own direct Met-binding site (MBS); Grb2 binding at pY1356 provides additional/indirect Gab1 recruitment (Birchmeier 2003 Fig 3 + Box 1 verified).
Koike 2025 footnote and body: added strain (C57BL/6J male), ages (8 wk young, 24 mo aged), n (4/group for qPCR), named the Mac_1 subpopulation specifically, added Cdkn1b suppression mechanism.
Ensembl ID: null → ENSG00000105976 (confirmed via UniProt P08581 REST cross-reference; gap tag removed).
Sequence length clarification: 1408 aa (Park 1987 ORF prediction) vs 1390 aa (UniProt canonical mature chain) — both now noted with context.

Downstream pages to check: dihexa (retraction context already noted in page; no new corrections needed from this pass).

[2026-05-09] verify — molecules/proteins/vegfr2.md

Pages verified: 1

molecules/proteins/vegfr2.md — corrections: 5 substantive; verified: true (partial scope — 4 primary PDFs read; 4 secondary sources not_oa/pending)

Sources checked:

Zeng 2001 (doi:10.1074/jbc.m103130200) — downloaded; full read
Bonner 2013 (doi:10.2337/db12-0354) — downloaded; full read
Smith 2020 eLife (doi:10.7554/eLife.54056) — downloaded; pp 1–8 read
Lian 2020 (doi:10.18632/aging.103834) — downloaded; pp 1–8 read
Terman 1991 (PMID:1656371) — abstract verified via PubMed efetch; no DOI (pre-DOI publication)
UniProt P35968 — identity fields confirmed via REST API (1356aa; HGNC:6307)
Shibuya 2011, Gustafsson 2005, Lloyd 2005, Hoier 2014 — not_oa; not verified
Zhao 2022 — pending at verification time; not verified

Corrections made (5 substantive):

Zeng 2001 phosphosite table: removed incorrect conflation of Tyr1054 with Tyr1059 as joint “activation loop” pair. Y1054F mutant showed no functional effect on proliferation or migration — now documented separately. Tyr1059 alone is responsible for MAPK/Ca²⁺/proliferation.
Zeng 2001 footnote: expanded with actual experimental system (EGFR-KDR chimera in primary HUVEC), per-site result granularity, and Y1054F no-op finding.
Bonner 2013 body + footnote: added cardiac muscle figure (~50% reduction vs ~60% skeletal), n=8/9, ex vivo isolated-muscle equivalence finding (implicating vascular delivery not myocyte signaling defect).
Lian 2020 direction corrected: wiki said “reversed by miR-25-3p inhibition” — actually reversed by miR-25-3p agomir (upregulation). miR-25-3p is downregulated in aging and promotes angiogenesis via suppression of TULA-2 → SYK Y323 → VEGFR-2 Y1175. Body and footnote both corrected.
Smith 2020 VE-cadherin section: added specificity that downstream VE-cadherin phosphosite is Y685 (not generic), pY658 was NOT significantly changed, and model was retinopathy (CNV + OIR). Footnote expanded with n, p-values, and per-model findings.

Downstream pages to check: vegfa, vegf-signaling may reference Zeng 2001 or Bonner 2013 claims.

[2026-05-09] verify — molecules/proteins/enos.md

Pages verified: 1 (partial — 5 of 8 cited sources verified; 4 closed-access)

Sources checked:

Förstermann & Sessa 2011 (doi:10.1093/eurheartj/ehr304) — downloaded; full read
Förstermann & Li 2010 (doi:10.1111/j.1476-5381.2010.01196.x) — downloaded; full read
Laufs 1998 (doi:10.1161/01.cir.97.12.1129) — already in archive; full read
Kim 2009 (doi:10.1152/japplphysiol.91393.2008) — downloaded; full read
Rinne 2013 (doi:10.1093/cvr/cvs335) — already in archive; full read
Dimmeler 1999 (doi:10.1038/21224) — closed-access (not_oa); not verifiable
Nisoli 2005 (doi:10.1126/science.1117728) — closed-access (not_oa); not verifiable
Lamas 1992 (doi:10.1073/pnas.89.14.6348) — closed-access (not_oa); not verifiable
Smith 2003 (doi:10.1042/bst0311447) — closed-access (not_oa); not verifiable

Corrections made (6 substantive):

BH4 cofactor description: “consumed stoichiometrically” → “oxidized to BH3 radical intermediate, recyclable or further degraded to BH2 under oxidative stress” (Förstermann 2011 Fig 2 caption)
Statin mechanism (pharmacology table + body text): “Rho/Rac1 → Akt axis” → “posttranscriptional mRNA stabilization via Rho GTPase geranylgeranylation inhibition” as the mechanism shown in Laufs 1998; PI3K/Akt is a secondary mechanism from review literature
Statin model (Laufs 1998 footnote): “HUVEC” → “human saphenous vein endothelial cells” (bovine aortic ECs for transfection studies); “rabbit carotid artery” confirmed
Rinne 2013 gap tag resolved: needs-doi-confirmation removed; DOI 10.1093/cvr/cvs335 confirmed; cAMP/PKA mechanistic intermediary removed as undemonstrated; model clarified (HUVECs, EA.hy926, C57BL/6N mice)
Tyr657 peroxynitrite claim: removed unsourced statement; replaced with confirmed BH4→BH2 amplification loop (Förstermann 2011)
Kim 2009 footnote: corrected from “mechanistic proof-of-concept for L-arginine competition” to accurate iNOS→SNO-Arg1→Arg1 activation→L-arginine depletion→eNOS uncoupling mechanism; added n=7, ages, ABH dose, and functional endpoint (Emax, PWV)

Unverifiable claims (closed-access sources):

~50% Ser1177 decline in aged rat aorta (Smith 2003) — tagged no-fulltext-access
Akt→eNOS Ser1177 phosphorylation (Dimmeler 1999) — consistent with Förstermann 2011 review confirmation but direct PDF unverifiable; tagged no-fulltext-access
CR→eNOS→PGC-1α→mtDNA axis in Nos3-null mice (Nisoli 2005) — tagged no-fulltext-access
eNOS cloning characterization (Lamas 1992) — tagged no-fulltext-access

verified: true (scope: partial — 5 sources PDF-verified; 4 closed-access)

Downstream propagation needed:

alpha-msh — cites Rinne 2013 and attributes “cAMP/PKA” intermediary; should be checked for same overclaim
pgc-1alpha — cites Nisoli 2005 via eNOS; Nisoli remains unverifiable (closed-access)
ampk — may cite eNOS Ser1177 as downstream target; verify attribution is correct

[2026-05-09] verify — biomarkers/glycanage-2017.md

Pages verified: 1

biomarkers/glycanage-2017.md — corrections: 9 substantive; verified: true (full primary-source scope for 6 of 8 cited papers; 2 secondary review papers and Mijakovac 2022 heritability not independently PDF-verified)

Sources checked (all PDFs downloaded and read):

Kristic 2014 (doi:10.1093/gerona/glt190) — downloaded; full read
Jurić 2020 (doi:10.18632/aging.104060) — downloaded; full read
Greto 2021 (doi:10.1038/s41366-021-00816-3) — downloaded; full read
Simunić-Briški 2024 (doi:10.1007/s10719-023-10144-5) — downloaded; full read
Rapcan 2024 (doi:10.1007/s11357-024-01239-4) — downloaded; full read
Murray 2023 (doi:10.1016/j.ebiom.2023.104692) — downloaded; partial read (abstract + methods + findings block)
Memarian 2023 (doi:10.1002/dmrr.3685) — downloaded; full abstract + methods read
Mijakovac 2022, Kristic/Lauc review 2022, Lauc 2024 expert opinion, Zhang 2026 review — not PDF-verified (review/secondary; lower claim weight)

Corrections made (9 substantive):

Cohort count: “5 European cohorts” → FOUR populations (Orkney n=2,035, TwinsUK n=1,261, Vis n=915, Korcula n=906). KORA was NOT in this study.
Cohort names: “KORA (Germany)” and “Vis-10000” removed; correct names used throughout.
Glycan peak identifiers: “FA2B, FA2G2, FA2BG2” → the model uses GP6, GP14, GP15 (chromatographic peak designations from the paper’s UPLC nomenclature).
Variance explained enriched: added R²=57.8% [54.1–61.3%] in Orkney training cohort; 41–50% in cross-population validation; prediction SD=9.7 yr.
Jurić 2020 — MAJOR factual error: wiki described “bilateral oophorectomy / surgical menopause at University of Colorado.” Source is a GnRH agonist (leuprolide) pharmacological suppression model in n=36 healthy premenopausal women (placebo n=21, estradiol n=15); NOT surgical menopause. n was stated in abstract (not “not confirmed”). IQR and p-value added.
Greto 2021 — n=37 now stated (was “n not confirmed”). Added specific glycan trait effect sizes (G0, G2, S1, CF with adj. p-values). Clarified that no numeric delta GlycanAge in years is reported; TwinsUK replication cohort (n=1,680) noted.
Simunić-Briški 2024 — Corrected from “three groups” to four groups (INACT, REC, ACT, PRO; n=276 total). Added PRO vs ACT FDR-status clarification (nominally significant overall; FDR-significant only in women). Men’s result correctly noted as not FDR-significant.
Murray 2023 — “~18–19 years” → precise paper figure “18.4–19.1 years”; added DYRK1A mechanism detail (increased unrepaired DNA damage + reduced LaminB1; reversed by hiPSC CRISPR/inhibitor).
Memarian 2023 — “incident nephropathy and macrovascular disease” → “both prevalent AND incident nephropathy and macrovascular disease”; sialylation-nephropathy association added; three cohort ns given (DiaGene n=1,815; GenodiabMar n=640; Hoorn DCS n=1,266).

Downstream propagation needed: glycanage-2017.md is a new page with no known citations from other wiki pages yet — no propagation required at this time.

[2026-05-09] verify — molecules/proteins/mc3r.md

Pages verified: 1

molecules/proteins/mc3r.md — corrections: 6 factual + footnote enrichment; verified: true (partial scope — Begriche/Butler 2011 closed-access unverified; canonical IDs not database-re-checked)

Sources checked: Lam 2021 (local PDF), Patel 2025 (local PDF), Roselli-Rehfuss 1993 (downloaded PMC47459), Mavrikaki 2016 (downloaded gold OA), Lonati 2020 (downloaded gold OA), Begriche 2011 (not_oa, not verified)

Corrections made (6 factual):

γ-MSH “Highest” affinity → “High/comparable to α-MSH”: Roselli-Rehfuss 1993 Ki γ₂-MSH (4.4 × 10⁻⁸ M) ≈ α-MSH (5.2 × 10⁻⁸ M); MC3R-preferring designation is distribution-based
Ligand-binding-selectivity paragraph updated with Ki values and corrected selectivity framing
Mouse puberty phenotype corrected: fasting abolished >2× oestrous cycle prolongation in WT but had no effect in Mc3r-null; males had 2-day delay; human 4.7-month menarche delay added with P-value and n
Patel 2025 mechanism corrected: LC3II activation + TFEB nuclear translocation (not generic “cAMP-dependent”); partial adiposity recovery qualifier added
IL-12 removed from MC3R-specific anti-inflammatory claims (unsupported by Lonati 2020 for MC3R); γ₂-MSH peritonitis evidence specified
Gαs-coupling MAPK/Ca²⁺ note clarified as post-1993 literature, not Roselli-Rehfuss attribution

Downstream propagation needed: none — mc3r.md not yet cited by other verified entity pages.

[2026-05-09] verify — pathways/camp-signaling.md

Pages verified: 1

pathways/camp-signaling.md — corrections: 7 substantive; verified: true (partial scope — Bos 2006, Ali 2020, Kang 2013 not_oa or download-failed and not independently verified; Yan 2007 cited secondarily via Enns 2010; Taylor 2013 and Kelly 2018 PMC full-text only)

Sources checked:

Enns 2010 (doi:10.18632/aging.100138) — local PDF (downloaded); full read; major misattribution found
Dehghan 2019 (doi:10.1038/s41467-019-12425-w) — local PDF (downloaded); full read; mechanism direction error found
Ye 2020 (doi:10.1111/acel.13055) — local PDF (downloaded); 6pp read; mechanism detail corrected
Wang 2021 (doi:10.1155/2021/6695613) — local PDF (downloaded); 8pp read; mouse strain corrected
Kelly 2018 (doi:10.1016/j.cellsig.2017.11.004) — PMC fallback (PMC5732030); download failed (green OA); AC-decline claim softened/corrected
Taylor 2013 (doi:10.1016/j.bbapap.2013.03.007) — PMC fallback (PMC3763834); download failed (green OA)
KEGG hsa04024 — confirmed “cAMP signaling pathway - Homo sapiens”
Reactome R-HSA-418555 — confirmed “G alpha (s) signalling events” (cAMP pathway)
Bos 2006, Ali 2020, Kang 2013 — not_oa or unverifiable; tagged unverified-source

Corrections:

Major attribution error — AC5-KO wrongly attributed to Enns 2010: The wiki stated “AC5-knockout mice have extended lifespan (~30% median lifespan increase) … [^enns2010].” Enns 2010 is about PKA RIIβ-KO and Cβ-KO mice. AC5-KO (~30% lifespan extension) is from Yan et al. 2007 (Cell 130:247-58), which Enns cites as ref [9]. Corrected to distinguish the two models; [^yan2007] footnote added.
Dehghan 2019 mechanism inverted: Wiki claimed hydralazine “inhibits [PKA] activity.” Paper shows hydralazine binds and stabilizes the PKA catalytic subunit causing regulatory-subunit dissociation, thereby activating PKA, which then activates SIRT1. Corrected throughout body and footnote.
Kelly 2018 AC-decline claim overstated: Wiki said Kelly 2018 “concluded that adenylyl cyclase activity in the hippocampus declines with age.” The review actually found ~half of studies showed age-related AC reduction and ~half showed no change; basal cAMP does not consistently fall. Corrected to accurately reflect mixed finding.
Kelly 2018 prefrontal/hippocampal conflation: Wiki grouped prefrontal cortex and hippocampus as having same cAMP decline. Kelly 2018 shows prefrontal cAMP paradoxically increases with age. Corrected with region-specific framing.
Ye 2020 mechanism detail: Wiki generically said “RIIα upregulation.” Paper specifies nuclear-fraction-specific PKAR2α elevation via PA28γ-20S proteasome complex disruption. Corrected in body and footnote.
Wang 2021 mouse strain: Footnote said “C57BL/6.” Paper uses CD1/ICR mice. Corrected in body and footnote; dose (60 mg/kg/d PTX i.p.) and group sizes added.
“Nuclear PGC-1α” removed: Wiki said Wang 2021 showed “nuclear PGC-1α” increase. Paper measures PGC-1α protein and mRNA levels, not nuclear localization. Corrected to “PGC-1α protein levels.”

Downstream pages to check (main agent):

Any page citing camp-signaling for the AC5-KO longevity claim should be checked for the wrong Enns 2010 attribution
sirtuin — if it references the hydralazine-PKA-SIRT1 axis, confirm the mechanism is PKA activation (not inhibition)
melanocortin-system — may cite camp-signaling for Gαs coupling claims; no corrections needed there

[2026-05-09] verify — pathways/pomc-processing.md

Pages verified: 1

pathways/pomc-processing.md — corrections: 7 substantive; verified: true (partial scope — Joshi 1995, Krude 1998, Jackson 1997 closed-access/not_oa; Bicknell 2008 and Hook 2009 download failed; UniProt IDs and Reactome ID verified via APIs; Nillni 2016 fully verified)

Sources checked:

Nillni 2016 JBC (doi:10.1074/jbc.m115.675264) — downloaded + full 16-page read; major attribution error found
Joshi 1995 Endocrinology (doi:10.1210/en.136.6.2721) — not_oa; verified against PubMed abstract (PMID 7750497)
Krude 1998 Nat Genet (doi:10.1038/509) — not_oa; verified against PubMed abstract (PMID 9620771)
UniProt P29120/P16519/P16870/P19021 — confirmed via REST API
Reactome IDs — corrected via ContentService API

Corrections:

Reactome frontmatter: R-HSA-264876-related → R-HSA-265301 (insulin processing → Corticotropin cleavage from POMC)
Nillni 2016 attribution (major): wiki claimed SIRT1→PC2 axis could shift POMC → α-MSH processing in arcuate neurons — paper studies CRH processing in PVN neurons exclusively; arcuate POMC neurons not directly studied; body section and footnote rewritten to accurately represent the paper’s actual subject and scope
Joshi 1995 ages: 3 months vs 12 months → 4–5 months vs 12–13 months
Joshi 1995 mechanism: general β-endorphin:α-MSH ratio → specific β-endorphin-(1-31) cleavage to -(1-27)/-(1-26); α-MSH processing not measured in this paper
Joshi 1995 quantification: PC2 increase now stated as ≥2-fold (double-label ISH), p < 0.05 (Northern blot)
Joshi 1995 furin: old mice showed significant furin mRNA reduction (not in wiki previously)
Reactome body section corrected; R-HSA-265301 hierarchy documented

Downstream pages to check (main agent):

molecules/proteins/pomc — may cite Joshi 1995 aging claims; check age values and mechanism description
melanocortin-system — if it references the SIRT1/POMC axis from Nillni 2016, attribution needs correction

[2026-05-09] verify — molecules/proteins/hgf.md

Pages verified: 1

molecules/proteins/hgf.md — corrections: 7 substantive; verified: true (partial scope)

Sources checked:

Koike 2025 Aging Cell (doi:10.1111/acel.70042) — local PDF; full read (17pp); primary source for Mac_1 aging mechanism
Birchmeier 2003 Nat Rev Mol Cell Biol (doi:10.1038/nrm1261) — local PDF; full read; domain architecture + Gab1 claims
Tu 2020 Aging Cell (doi:10.1111/acel.13116) — downloaded + 5pp read; Werner syndrome MSC claims confirmed
Rodgers 2017 Cell Rep (doi:10.1016/j.celrep.2017.03.066) — abstract only via PubMed efetch (gold OA but PDF download failed); G_Alert + HGFA claims confirmed from abstract
Nakamura 1989 Nature (doi:10.1038/342440a0) — not_oa; unverifiable; ~1 ng/ml claim tagged no-fulltext-access
Trusolino 2010 Nat Rev Mol Cell Biol (doi:10.1038/nrm3012) — bronze OA but download failed; MET signalling description tagged no-fulltext-access
Canonical IDs (UniProt P14210, NCBI Gene 3082, HGNC:4893, ENSG00000019991) — re-confirmed against UniProt REST API and NCBI Gene

Corrections:

Koike 2025 — macrophage subpopulation named specifically as “Mac_1” (was generic “a specific macrophage subpopulation”)
Koike 2025 — strain corrected to C57BL/6J (was C57BL/6); male mice only
Koike 2025 — Cdkn1b suppression mechanism added explicitly (key mechanistic finding, was absent)
Koike 2025 — HGF rescue dose added (0.5 μg/day every other day); n confirmed (n=4 qPCR; n=3 in vivo)
Koike 2025 — “age-related depletion” softened to “reduced HGF expression in macrophages + reduced Mac_1 proliferative capacity” (paper shows reduced function, not elimination)
Birchmeier 2003 — “N-terminal PAN/apple domain” corrected to “N-terminal domain (N)” (Birchmeier 2003 does not use PAN/apple nomenclature)
Birchmeier 2003 — Gab1 unsourced tag removed; claim is explicitly supported in Birchmeier 2003 (added citation)
Rodgers 2017 — FAPs added alongside MuSCs as G_Alert targets (per abstract)
Tu 2020 — footnote updated: iPSC-derived MSC model clarified (not direct MSC correction); n confirmed (2 corrected clones)

Downstream pages to check (main agent):

None currently link to hgf.md via study footnotes (page is standalone protein page)

[2026-05-09] verify — molecules/compounds/colchicine.md

Pages verified: 1

molecules/compounds/colchicine.md — corrections: 5 substantive (see below); verified: true (partial scope — PK table and canonical identity fields not re-verified against package insert/DrugBank)

Sources checked:

COLCOT 2019 (doi:10.1056/NEJMoa1912388) — downloaded + full read; safety data corrections
LoDoCo2 2020 (doi:10.1056/NEJMoa2021372) — downloaded + full read; GI/pneumonia safety corrections
COPPS-2 2014 (doi:10.1001/jama.2014.11026) — downloaded + full read; p-value correction
Toldo 2022 (doi:10.1016/j.pharmthera.2021.108053) — local PDF; full §5.5 read; mechanism claims confirmed

Corrections:

COLCOT safety: GI AE rate corrected (17.5% vs 17.6%, NS — not “7.7% vs 5.3%”); non-CV death corrected to 23 vs 20 patients (not a percentage); pneumonia 0.9% vs 0.4% (p=0.03) confirmed but correctly attributed to COLCOT
LoDoCo2 safety: Wrong GI figure removed (“9.7% vs 6.4%”); pneumonia signal corrected (LoDoCo2 showed NS, 1.7% vs 2.0%; the 0.9%/0.4% p=0.03 figure was misattributed from COLCOT); noncardiovascular death HR 1.51 (95% CI 0.99–2.31) added
COPPS-2 p-value: 0.002 → 0.046 (log-rank); RR expression removed, replaced with absolute difference + 95% CI per source
Safety profile table: Pneumonia signal correctly attributed to COLCOT (not LoDoCo2)
clinical-trials-active: 7 → 68 (re-queried ClinicalTrials.gov v2 API 2026-05-09)

Downstream pages to check (main agent):

None currently link directly to colchicine.md via study footnotes (compound page is standalone)

[2026-05-09] verify — molecules/proteins/mc5r.md

Pages verified: 1

molecules/proteins/mc5r.md — corrections: 7 substantive; verified: true (partial scope — Ng 2021, Yang 2013, Rodrigues 2017, Zhang 2006, Janner 1998 not fully verified due to access issues)

Sources checked:

Chen 1997 Cell (doi:10.1016/s0092-8674(00)80467-5) — local PDF; full read; multiple corrections
Ji 2022 IJMS (doi:10.3390/ijms23158727) — downloaded + full read
Rodrigues 2009 Mol Cell Endocrinol (doi:10.1016/j.mce.2009.01.014) — downloaded + full read
Thiboutot 2000 J Invest Dermatol (doi:10.1046/j.1523-1747.2000.00094.x) — downloaded + full read
McDonald 2021 Front Immunol (doi:10.3389/fimmu.2021.742154) — downloaded + full read
Xu 2020 Cell Mol Life Sci (doi:10.1007/s00018-020-03511-0) — downloaded + full read
Ng 2021 (doi:10.1080/09273948.2020.1849735) — download failed (no OA URL); tagged no-fulltext-access

Key corrections: Ensembl ID corrected (ENSG00000198822→ENSG00000176136; prior ID was GRM3); sweat/eccrine row removed from KO table (not in source); sebum/water quantified; lacrimal EC₅₀ added; Harderian strain-specificity noted; PI3K/ERK made Akt-independent; MRAP description made specific; ligand affinity order corrected (ACTH=β-MSH); Thiboutot human-preputial-cells claim corrected to rat

[2026-05-09] verify — processes/melanogenesis.md

Pages verified: 1

processes/melanogenesis.md — corrections: 3 factual + multiple no-fulltext-access tags added; verified: true (partial scope — all five primary sources closed-access; Nishimura 2005 + Kovacs 2010 partially verifiable via Crossref/PubMed abstracts)

Sources checked:

Nishimura 2005 Science (doi:10.1126/science.1099593) — not_oa; Crossref abstract retrieved; mechanisms partially verified
Slominski 2004 Physiol Rev (doi:10.1152/physrev.00044.2003) — not_oa; PubMed abstract retrieved; biochemical cascade detail unverifiable
Costin 2007 FASEB J (doi:10.1096/fj.06-6649rev) — not_oa; abstract only; melanosome stage/transfer detail unverifiable
Kovacs 2010 Br J Dermatol (doi:10.1111/j.1365-2133.2010.09946.x) — not_oa; PubMed abstract retrieved; growth factors verified
Wan 2011 Mol Cell Biochem (doi:10.1007/s11010-011-0823-4) — not_oa; abstract only; MITF phosphorylation sites unverifiable
Zhang 2023 J Cosmet Dermatol (doi:10.1111/jocd.15652) — hybrid OA; download failed; PubMed abstract retrieved

Corrections made:

bFGF → KGF: Kovacs 2010 examined SCF, HGF, and KGF (keratinocyte growth factor), NOT bFGF; wiki incorrectly listed bFGF in age-spot mechanism section and footnote
“Senescent fibroblasts” → “Photoaged fibroblasts”: Kovacs 2010 abstract describes “photoaged fibroblasts”; senescent characterization not stated in abstract no-fulltext-access
McSC depletion mechanisms: removed “DNA damage-driven apoptosis” and “incomplete Notch-mediated self-renewal” as attributed mechanisms for Nishimura 2005 — these are secondary literature claims not present in Nishimura 2005 abstract; Nishimura 2005 demonstrates only (1) ectopic differentiation and (2) BCL2-deficiency-driven apoptosis selectively of McSCs during niche dormancy entry; Notch/DNA-damage mechanisms tagged needs-replication
Added no-fulltext-access tags throughout for enzymatic cascade, melanosome stage morphology, MITF phosphorylation sites (Ser73, Ser133, Ser307, Ser69, Ser409) — unverifiable from closed-access PDFs
Updated footnotes to accurately reflect what each abstract confirms vs. does not confirm
Cross-reference note for cellular-senescence updated to reflect uncertainty about fibroblast senescence in Kovacs 2010

Downstream pages to check (main agent):

molecules/proteins/bcl2.md — if it cites Nishimura 2005 for McSC protection claim
hallmarks/stem-cell-exhaustion.md — if it cites McSC ectopic differentiation mechanism from Nishimura 2005 (check mechanism description matches corrected version)
hallmarks/cellular-senescence.md — if it cites Kovacs 2010; bFGF→KGF correction needed there too

[2026-05-09] verify — pathways/trkb-pathway.md

Pages verified: 1

pathways/trkb-pathway.md — corrections: 5; verified: true (partial scope — Webster 2006 and Minichiello 2009 closed-access)

Sources checked:

Jang 2010 PNAS (doi:10.1073/pnas.0913572107) — downloaded (PMC2823863) + full PDF read; selectivity claim verified (TrkB not TrkA/TrkC); ECD CC-2 binding domain verified; Kd ≈ 320 nM confirmed; i.p. 5 mg/kg dose confirmed; in-vivo models are mouse-only (no rats); no aging experiment in paper
Reichardt 2006 Phil Trans R Soc B (doi:10.1098/rstb.2006.1894) — downloaded (PMC1664664) + full PDF read; three-arm signaling architecture verified; Y490/Y785 TrkA numbering confirmed (rat convention); TrkB.T1 dominant-negative mechanism described as inhibiting productive dimerization (not primarily BDNF sequestration); FRS2 coupling to PI3K via juxtamembrane pY confirmed
Webster 2006 Gene Expression Patterns (doi:10.1016/j.modgep.2006.03.009) — not_oa; TrkB.FL:T1 aging ratio claims tagged no-fulltext-access
Minichiello 2009 Nat Rev Neurosci (doi:10.1038/nrn2738) — not_oa (PMID: 19927149); Y515F/Y816F knock-in phenotype claims tagged no-fulltext-access
NTRK2 phosphotyrosine numbering: rat P15209 (Y515/Y705/Y706/Y816) vs human Q16620 (Y516/Y702/Y706/Y707/Y817) — documented in body; rat convention confirmed as correct for cited experimental literature
Reactome R-HSA-9006115 confirmed as “Signaling by NTRK2 (TRKB)” — correct
KEGG hsa04722 confirmed as “Neurotrophin signaling pathway” — correct (pan-neurotrophin, not TrkB-specific; acknowledged in wiki)

Corrections made:

Jang 2010 footnote model: “rat/mouse; stroke + PD models” → “mouse only (TrkB F616A knockin + C57BL/6; no aging arm)”
Jang 2010 footnote: removed “locally pending” → “verified against local PDF (PMC2823863)”
Jang 2010 footnote: added Kd ≈ 320 nM, n=4–12 mice/group, specific model details
Reichardt 2006 footnote: removed “locally pending” → “verified against local PDF (PMC1664664)”
Pharmacology table 7,8-DHF row: removed [^jang2010] from “rescues memory in aged rodents” (Jang 2010 has no aging experiment); aged-rodent memory rescue attributed correctly to [^zeng2012jnc] only; added Kd ≈ 320 nM with correct attribution to jang2010
Isoform table: TrkB.T1 primary function corrected from “Sequesters BDNF” → “Inhibits productive dimerization with TrkB.FL (dominant-negative)” per Reichardt 2006
Phosphotyrosine sites: added human UniProt Q16620 numbering in parentheses throughout; added explanatory note on rat vs human convention
Added no-fulltext-access tags to Webster 2006 and Minichiello 2009 footnotes
Updated page banner to partial-verification notice (replacing auto-extraction warning)

Downstream pages to check (main agent):

molecules/compounds/ pages for 7,8-DHF / semax / dihexa — pharmacology table corrections may propagate
molecules/proteins/bdnf.md — if it cites jang2010 for aged-rodent rescue
molecules/proteins/ntrk2.md (if seeded) — Y-site numbering

[2026-05-09] verify — pathways/spm-pathway.md

Pages verified: 1

pathways/spm-pathway.md — corrections: 4; verified: true (partial scope — Chiang 2017 and Serhan 2014 Nature closed-access; Doyle 2018 closed-access; WikiPathways WP5155 not re-verifiable via current API)

Sources checked:

Serhan 2018 JCI (doi:10.1172/JCI97943) — local PDF read end-to-end; primary source for biosynthetic routes, receptor pharmacology, four-family framework; confirmed
Dalli 2017 Front Immunol (doi:10.3389/fimmu.2017.01400) — downloaded + full PDF read; maresin tissue distribution claims verified; footnote description corrected
Ponce 2022 Front Aging Neurosci (doi:10.3389/fnagi.2022.780811) — downloaded + read; neurodegenerative disease framing consistent with wiki
Mehrotra 2022 Nat Rev Drug Discov (doi:10.1038/s41573-022-00470-y) — downloaded + read; efferocytosis/SPM pharmacology framing consistent
Chiang 2017 Mol Aspects Med (doi:10.1016/j.mam.2017.03.005) — not_oa (download failed); receptor pharmacology claims partially unverified
Serhan 2014 Nature (doi:10.1038/nature13479) — not_oa (download failed); overview framing unverified against full PDF; Crossref metadata confirmed (title, journal, year correct)
Doyle 2018 Semin Immunol (doi:10.1016/j.smim.2018.09.002) — not_oa; aging-decline claims unverified
zhang2023 Hypertension (doi:10.1161/HYPERTENSIONAHA.123.21348) — not_oa; study design confirmed as mouse model from title
romano2006 (doi:10.2174/187152806776383152) — Crossref title confirmed correct; local status not confirmed
Reactome R-HSA-9018678, R-HSA-9018676, R-HSA-9020265 — all confirmed correct via REST API
WikiPathways WP5155 — current API endpoints non-functional; not re-verifiable

Corrections:

GPR32 receptor ligand list: “RvD1, RvD2, RvD3” → “RvD1, AT-RvD1, RvD3, RvD5”. Per Serhan 2018 JCI p.2662 and Dalli 2017 Table 1, GPR32/DRV1 binds RvD1/AT-RvD1/RvD3/RvD5; RvD2 is a ligand for GPR18/DRV2, not GPR32. RvD2 was incorrectly attributed to GPR32.
Maresins tissue distribution: “resolving exudates and bone marrow macrophages” → “resolving exudates, mouse infectious exudates, human macrophages, spleens, plasma.” “Bone marrow macrophages” claim traces to Abdoulnour 2014 (not locally verified); noted with source and gap tag.
zhang2023 body claim: “RvE1 serum levels significantly decreased in hypertensive patients” → “RvE1 levels reduced in Angiotensin II–induced hypertensive mouse model.” Paper title specifies mouse model; human patient claim unsupported from available metadata.
zhang2023 footnote study design: “in-vivo + human observational” → “in-vivo (mouse).” Consistent with paper title.
Dalli 2017 footnote description: “maresins and macrophage function in resolution; efferocytosis and tissue repair” → full scope description covering all SPM families in macrophage biology, microvesicle regulation, maresin bioactive metabolome.

Downstream pages to check:

fpr2 — cites GPR32 receptor pharmacology; verify it does not also list RvD2 as GPR32 ligand
gpr32 — stub page; if seeded, will need corrected ligand list (RvD1/AT-RvD1/RvD3/RvD5, not RvD2)
chemr23 — stub page; RvE1/RvE2 receptor confirmed correct, no correction needed

[2026-05-09] verify — pathways/neurotrophin-signaling.md

Pages verified: 1

pathways/neurotrophin-signaling.md — corrections: 4; verified: true (partial scope — Huang & Reichardt 2001/2003 and Bibel & Barde 2000 not verified against primary source PDFs; download failed or closed-access)

Sources checked:

Nykjaer 2004 (doi:10.1038/nature02319) — local PDF read; claims verified
Latina 2017 (doi:10.3389/fncel.2017.00068) — downloaded + full PDF read; major misattribution found and corrected
Huang & Reichardt 2001 (doi:10.1146/annurev.neuro.24.1.677) — green OA, download failed; unverified
Bibel & Barde 2000 (doi:10.1101/gad.841400) — diamond OA, download failed (HTTP 520); unverified
Huang & Reichardt 2003 (doi:10.1146/annurev.biochem.72.121801.161629) — closed-access (not_oa); unverified
Bregman-Yemini 2024 (doi:10.1016/j.arr.2024.102349) — closed-access (not_oa); gap tag maintained
Wrann 2013, Sleiman 2016 — framing spot-checked vs bdnf page; consistent; no corrections needed
KEGG hsa04722 — confirmed correct (Neurotrophin signaling pathway, Homo sapiens)
Reactome R-HSA-166520 — confirmed correct (Signaling by NTRKs, Homo sapiens)
WikiPathways WP4149 — confirmed WRONG candidate (White fat cell differentiation); corrected in body
WikiPathways WP2380 — confirmed WRONG candidate for this page (BDNF-specific only); corrected in body

Corrections:

p75NTR apoptotic cascade: JNK→p53→BAX not supported by Nykjaer 2004 alone. Statement qualified — Nykjaer 2004 establishes the sortilin/p75NTR complex requirement; downstream cascade specifics come from secondary literature. Sortilin binding affinities added from SPR data (Kd ~5 nM for proNGF, ~87 nM for NGF).
Latina 2017 body claim: “TrkA protein levels decline in MCI and AD before significant cholinergic neuron loss occurs” misattributed to Latina 2017. That paper is an in vitro rat culture study reporting presynaptic dysfunction upon NGF withdrawal, not a human MCI/AD cohort study. Body corrected; actual Latina 2017 finding (reversible presynaptic dysfunction in E17/18 rat cultures) now stated.
Latina 2017 footnote model: “primary rat basal forebrain cholinergic neurons” → “embryonic day 17/18 Wistar rat septo-hippocampal primary cultures enriched for cholinergic neurons.” Key quantitative data (n, reversibility window, TrkA-dependence) added.
WikiPathways gap note: WP4149 and WP2380 named as unconfirmed candidates → corrected to state both are confirmed wrong. WP4149 = “White fat cell differentiation”; WP2380 = “BDNF signaling” (BDNF-specific). Canonical null maintained.

Downstream pages to check:

None identified (this pathway page does not currently propagate specific study-derived quantitative claims to other pages; BDNF page is the downstream anchor and its Wrann/Sleiman framing is consistent)

[2026-05-09] verify — molecules/proteins/pomc.md

Pages verified: 1

molecules/proteins/pomc.md — corrections: 6; verified: true (partial scope — Bicknell 2008 and Seidah 1999 PDFs inaccessible; Coll 2004, Kühnen 2019, Nillni 2016 not_oa; Krude 1998 not_oa but verified via PubMed abstract)

Sources checked:

UniProt P01189 — REST API; canonical IDs and peptide chain annotations confirmed
Krude 1998 (doi:10.1038/509, PMID 9620771) — not_oa; verified via PubMed efetch abstract
Biebermann 2006 (doi:10.1016/j.cmet.2006.01.007) — full PDF downloaded and read (bronze OA)
Sadagurski 2015 (doi:10.14814/phy2.12379) — full PDF downloaded and read (gold OA/PMC)
Bicknell 2008, Seidah 1999, Coll 2004, Kühnen 2019, Nillni 2016 — not verified (not_oa or download failure)

Corrections:

Krude 1998 genotype: “homozygous POMC loss-of-function mutations” → Patient 1 compound heterozygote (exon-3: G7013T + C7133Δ); Patient 2 homozygous (exon-2: C3804A). Body + footnote corrected.
JP residue positions wrong: 103–137 → 105–134 (UniProt P01189 “Potential peptide”); Ser168 phospho claim removed.
γ-endorphin table row: flagged as not an annotated UniProt chain feature (#gap/needs-canonical-id).
Sadagurski 2015 footnote misattributed lifespan extension to this paper (belongs to Sun et al. 2009); footnote rewritten with actual findings.
Sadagurski 2015 body text: mechanism description updated to match paper (Ikbkb/TNF-α/Iba1+ microglia reduced by CL intervention).
Biebermann 2006 footnote: added n (722 obese/1270 controls, 2 vs 0 carriers), quantitative in-vitro data (100-fold Ki, 50-fold cAMP), in-vivo ICV rat confirmation; design label corrected.

Downstream pages to check (main agent):

pathways/melanocortin-system.md — check if Krude 1998 genotype correction needs propagation.
Any future studies/krude-1998-pomc-lof.md study page must use compound-heterozygote framing for Patient 1.

[2026-05-09] verify — molecules/proteins/il-6.md

Pages verified: 1

molecules/proteins/il-6.md — corrections: 5 factual changes (see details below); verified: true (partial scope — Schreiber 2021 abstract-only; Franceschi 2000 and Pedersen 2007 closed-access)

Sources checked:

Swerdlow 2012 (doi:10.1016/S0140-6736(12)60110-X) — local PDF, verified end-to-end
Rose-John 2012 (doi:10.7150/ijbs.4989) — downloaded and verified end-to-end
Heink 2017 (doi:10.1038/ni.3632) — downloaded and verified end-to-end
Schreiber 2021 (doi:10.1053/j.gastro.2021.02.062) — hybrid OA, PDF inaccessible; abstract confirmed 44%/19% figures and n=16
Franceschi 2000 (doi:10.1111/j.1749-6632.2000.tb06651.x) — not_oa, unverifiable
Pedersen 2007 (doi:10.1016/j.tips.2007.02.002) — not_oa, unverifiable

Corrections made:

CHD OR: “~30% lower odds” → OR 0.95 (95% CI 0.93–0.97), p=1.53×10⁻⁵ per allele (Swerdlow 2012)
CHD case count: “~40,000 cases” → 25,458 cases / 100,740 controls (34 studies); 133,449 individuals total across 40 studies for biomarker analyses
MR instrument: “rs2228145” as primary instrument → rs7529229 is the lead analytic SNP; rs8192284/rs2228145 p.Asp358Ala is the functional correlate in LD (r²=0.92)
sIL-6R plasma level: “300–400 ng/mL” → ~75 ng/mL for sIL-6R; 250–400 ng/mL is for sgp130 (Rose-John 2012)
ADAM10/ADAM17 shedding: “ADAM10/ADAM17” → ADAM17 is the primary sheddase; ADAM10 role is secondary/conditional (Rose-John 2012)
Trans-presentation DC subset: added precision — CD11b+Sirpα+CD103⁻SiglecH⁻ (cDC2) subset; IL-6 loaded intracellularly; anti-IL-6 Abs fail to block (Heink 2017)

Pages unverifiable (closed-access):

Franceschi 2000 — inflammaging concept framing unverifiable
Pedersen 2007 — myokine framing unverifiable

Downstream pages potentially needing updates:

Any study page for Swerdlow 2012 if one exists
sasp (if it cites IL-6 MR figures)
chronic-inflammation (if it cites CHD OR or rs2228145 instrument)

[2026-05-09] verify — melanocortin-system.md (MOC)

Pages verified: 1 (navigational MOC)

pathways/melanocortin-system.md — corrections: 4; verified: true (partial scope — review PDFs not read end-to-end; MOC makes no primary quantitative claims)

Sources checked:

Reactome R-HSA-388601 — ContentService REST API; confirmed DefinedSet “Melanocortin receptors [plasma membrane]”; stId valid
Reactome R-HSA-500792 — confirmed as “GPCR ligand binding” (not “GPCR downstream signaling” as original text stated)
Reactome R-HSA-388596 — confirmed as “Melanocortin receptors bind melanocortins” (not “melanocortins bind MC receptors”)
Reactome R-HSA-388605 — confirmed as “POMC(138-176) binds MC2R” (not “ACTH binds MC2R” — substantively same, nomenclature updated)
Girardet 2014 (doi:10.1016/j.bbadis.2013.05.004) — DOI lookup confirmed title match; OA (green); PDF not downloaded
Quarta 2021 (doi:10.1038/s42255-021-00345-3) — DOI lookup confirmed title match; OA (green); PDF not downloaded
Dores 2016 (doi:10.1530/jme-15-0292) — DOI lookup confirmed title match; OA (bronze); PDF not downloaded
Cross-links checked: alpha-msh.md, mc1r.md, mc4r.md, kpv.md, melanotan-ii.md, pomc.md confirmed to exist; agrp.md, asip.md, mc2r.md, mc3r.md, mc5r.md, camp-signaling.md, hpa-axis.md, melanogenesis.md, hypothalamic-arcuate-circuit.md confirmed absent (correctly flagged as stubs)

Corrections:

Reactome parent pathway name wrong: “GPCR downstream signaling” → “GPCR ligand binding” (R-HSA-500792 confirmed via REST)
R-HSA-388601 schema type wrong: “parent reaction set” → “DefinedSet” (entity collection, not a Pathway/Reaction)
Sibling reaction names updated: “melanocortins bind MC receptors” → “Melanocortin receptors bind melanocortins”; “ACTH binds MC2R” → “POMC(138-176) binds MC2R”
POMC page status corrected: listed as “stub; protein page not yet seeded” — pomc.md exists and was seeded 2026-05-09; Cross-references and Implicit stubs sections and stub note all corrected

Downstream pages to check (main agent):

None required — this is a navigational MOC; the Reactome ID corrections are internal to this page only.

[2026-05-09] verify — mc4r.md

Pages verified: 1

molecules/proteins/mc4r.md — corrections: 6 (see below); verified: true (partial scope — Clément 2020 not_oa)

Sources checked:

UniProt P32245 — REST API; canonical IDs confirmed (NCBI 4160, HGNC 6932, Ensembl ENSG00000166603, length 332 aa, MW 36,943 Da)
Farooqi 2003 (doi:10.1056/NEJMoa022050) — full PDF downloaded and read (NEJM)
Lotta 2019 (doi:10.1016/j.cell.2019.03.044) — full PDF downloaded and read (Cell)
Collet 2017 (doi:10.1016/j.molmet.2017.06.015) — full PDF downloaded and read (Mol Metab)
Baldini 2019 (doi:10.1530/JOE-18-0596) — full PDF downloaded and read (J Endocrinol)
Haqq 2022 (doi:10.1016/S2213-8587(22)00277-7) — local PDF download failed; PMC9847480 full text verified via WebFetch
Clément 2020 (doi:10.1016/S2213-8587(20)30364-8) — NOT VERIFIED: closed-access (not_oa per a local paper archive)
FDA NDA213793 — openFDA API; approval dates confirmed

Corrections:

Haqq 2022 n wrong: wiki had “n=44 BBS + 3 Alström” → actual n=32 BBS + 6 Alström = 38 total enrolled. Corrected in body and footnote.
Haqq 2022 endpoint precision: added 95% CI (16.7–51.4%), p=0.0006, 52-week timepoint, and “all responders had BBS” qualifier.
FDA approval structure wrong: LEPR deficiency was in original Nov 25, 2020 approval (ORIG-1), not a separate “extended 2020” — corrected. BBS added June 16, 2022 (SUPPL-1) — date now precise.
Collet 2017 mischaracterised as “small case series” — was actually a randomised double-blind placebo-controlled Phase 1b RCT (n=6 MC4R-het + n=2 placebo + 41 obese controls, 28 days). Body and footnote corrected.
Lotta 2019 n inflated: “~550,000 participants” → ~452,300 UK Biobank (BMI n=450,708); “~0.5 million including replication cohorts.” Added precise ORs and β-arrestin R²=0.88 finding. Corrected in body, MR section, and footnote.
Clément 2020 unverifiable status: flagged with no-fulltext-access in body near claims and in footnote.

Pages unverifiable (closed-access):

Clément 2020 (Lancet DE) — not_oa; no-fulltext-access added. Claims (80% POMC / ~45% LEPR responder rates) sourced from secondary literature only.

Downstream pages to check (main agent):

molecules/compounds/setmelanotide.md (implicit stub) — if seeded, check Haqq 2022 n and Clément 2020 responder rates propagated correctly.
Any MOC or intervention page citing setmelanotide Phase 3 outcome figures.

[2026-05-09] verify — mc1r.md

Pages verified: 1

molecules/proteins/mc1r.md — corrections: 4 (see below)

Sources checked:

Gan-Or 2016 (doi:10.1016/j.neurobiolaging.2016.03.029) — PMC full text + PubMed abstract; green OA (PMC4892956)
Feng 2015 (doi:10.1186/s40035-015-0043-z) — full PDF downloaded and read
Suzuki 1996 (doi:10.1210/endo.137.5.8612494) — abstract only; bronze OA, download failed
Kadekaro 2012 (doi:10.1158/1541-7786.MCR-11-0436) — abstract only; closed-access (not_oa)
UniProt Q01726, NCBI Gene API, FDA OpenFDA — canonical IDs + approval date confirmed

Corrections:

Gan-Or 2016 conditioning claim — “no significant independent association after conditioning on melanoma risk alleles” is factually wrong; actual result: no per-cohort significance; meta-analysis OR=1.22 (p=0.03, CI 1.02–1.47) with heterogeneity; OR=1.11 (p=0.27) after removing heterogeneous study. Melanoma-allele conditioning not performed in this paper.
Suzuki 1996 Kd values removed — abstract reports relative-affinity order only (α-MSH = ACTH > β-MSH > γ-MSH), not sub-nanomolar Kd values; “sub-nanomolar Kd” claim removed, gap-tagged.
Feng 2015 footnote corrected — was labelled primary source for R160W; is a review synthesizing Tell-Marti 2015 and others; footnote and body now correctly identify it as review-level.
Kadekaro 2012 footnote updated — mechanism confirmed from abstract (p53 Ser15 → OGG1 + APE-1/Ref-1); closed-access status flagged with no-fulltext-access.

Canonical IDs re-verified: UniProt Q01726, NCBI 4157, HGNC:6929, Ensembl ENSG00000258839 — all match. FDA ORIG approval 2019-10-08 confirmed.

Pages unverifiable:

Kadekaro 2012 — not_oa; abstract only. no-fulltext-access added to footnote.
Suzuki 1996 — bronze OA, download failed; abstract only. no-fulltext-access added to footnote.

Downstream pages to check (main agent):

molecules/compounds/melanotan-ii.md — may cite MC1R Kd claims; verify “sub-nanomolar” not propagated.
molecules/proteins/alpha-msh.md — cites Suzuki 1996; same abstract-only limitation applies.

[2026-05-09] verify — setmelanotide.md

Pages verified: 1

molecules/compounds/setmelanotide.md — corrections: 4 (see below); verified: true (partial scope — Haqq 2022 and Clément 2020 PDFs unavailable; VENTURE 2025 abstract-only)

Sources checked:

Collet 2017 (doi:10.1016/j.molmet.2017.06.015) — full local PDF read end-to-end; all quantitative claims verified
VENTURE 2025 / Argente et al. (doi:10.1016/S2213-8587(24)00273-0; PMID 39549719) — verified via PubMed efetch (full abstract); published Lancet DE 2025;13(1):29-37
PubChem CID 11993702 — REST API; molecular formula C49H68N18O9S2, MW 1117.3 Da, InChIKey HDHDTKMUACZDAA-PHNIDTBTSA-N all confirmed
Haqq 2022 (doi:10.1016/S2213-8587(22)00277-7) — local PDF download failed; PMC fetch also failed; body claims accepted from R35-Stage2 MC4R verifier corrections (n=38, primary endpoint 32.3%, CI/p confirmed)
Clément 2020 (doi:10.1016/S2213-8587(20)30364-8) — not_oa; body claims not independently verified; accepted from R35-Stage2 corrections
Markham 2021 (doi:10.1007/s40265-021-01470-9) — not_oa; not verified; claims sourced from secondary literature

Corrections:

Collet 2017 n mischaracterised — wiki stated “n=6 setmelanotide + n=2 placebo in MC4R-deficient obese subjects, plus n=41 common obese controls.” Actual: n=49 total across 5 cohorts (41M+8F), of whom n=8 were MC4R heterozygous carriers (6 setmelanotide + 2 placebo) and n=41 were obese controls enrolled across all cohorts. Table 1 shows 5+3=8 obese controls in the specific weight-comparison subset. Full design description corrected.
Collet 2017 weight-loss attribution imprecise — “−3.48 kg… in the active-treatment arm” now correctly attributed to the MC4R heterozygous carrier arm (95% CI −4.99 to −1.96); obese control arm result (−3.07 kg, 95% CI −4.11 to −2.04, p<0.0001) added.
Collet 2017 dose added — 0.01 mg/kg/24h subcutaneous continuous infusion for 28 days (distinct from approved Imcivree 1–3 mg/day).
VENTURE 2025 status wrong — wiki said “results pending full publication.” Paper is fully published: Lancet DE 2025;13(1):29-37. Body and footnote corrected with n=12, co-primary endpoints, BMI outcomes, AE profile, and DOI.

Pages unverifiable:

Clément 2020 — not_oa; claims (80% POMC / 45% LEPR responder rates) accepted from R35-Stage2 corrections only
Haqq 2022 — download failed; PMC fetch also failed; claims accepted from R35-Stage2 corrections only
Markham 2021 — not_oa; not verified

Downstream pages to check (main agent):

None required — setmelanotide.md is the terminal compound page; no other wiki pages currently cite these study-level claims via setmelanotide footnotes. MC4R page corrections already propagated in R35-Stage2.

[2026-05-09] verify — mots-c-peptide.md

Pages verified: 1

molecules/proteins/mots-c-peptide.md — corrections: 4 (see below)

Sources checked:

Lee 2015 (doi:10.1016/j.cmet.2015.02.009) — PDF verified; sequence, mechanism, dose confirmed
Reynolds 2021 (doi:10.1038/s41467-020-20790-0) — PDF verified; n=10, 11.9-fold, late-life protocol confirmed
Miller 2022 (doi:10.1172/JCI158449) — PDF verified; 8 MDP count confirmed
Zempo 2021 (doi:10.18632/aging.202529) — PDF verified; centenarian null finding confirmed; OR confirmed
D’Souza 2020 (doi:10.18632/aging.102944) — PDF newly downloaded and verified; n=26/group confirmed
UniProt A0A0C5B5G6 — re-checked mutagenesis data

Corrections:

Kinesio-genomic p-value: body text and footnote corrected from “p=0.01” to “p=0.014” (p<0.05, not p<0.01 as previously stated)
UniProt mutagenesis residues 13-16: corrected from “nuclear localisation AND DNA binding” to “DNA binding ONLY” (nuclear localisation is not abolished by 13-16 AAAA substitution)
D’Souza 2020 n: added “n=26 per group” (was vaguely “n cross-sectional”); added plasma decline quantification (~11% middle, ~21% older)
D’Souza 2020 footnote: upgraded from “download pending” to verified with full citation data

Pages unverifiable: none (D’Souza 2020 downloaded during this session)

Downstream pages to check: molecules/compounds/mots-c.md — already verified (R36); the corrected Zempo p-value (p=0.014 not p<0.01) matches the compound page’s framing which does not state the p-value numerically. No propagation needed.

[2026-05-09] R35-Stage2 + R36-followup close — biomarkers + parent proteins/receptors

User-requested round following the R36 peptide-cluster close: seed and verify hallmark-tracking biomarker pages (R35 Stage 2) plus the parent proteins/receptors surfaced as implicit stubs in R36. 18 atomic content pages seeded + 16 verified to date (3 wave-3 pages still in seeder dispatch at log-write time). Round mirrored the R36 cadence: parallel seeder/verifier waves of 8 + 8 + 3.

Pages added (wave 1 + 2; wave 3 still in flight)

Biomarker pages (10):

biomarkers/gdf-15-biomarker.md
biomarkers/il-6-biomarker.md
biomarkers/nfl-biomarker.md
biomarkers/chip-clonal-hematopoiesis-biomarker.md
biomarkers/igf-1-biomarker.md
biomarkers/nlr-biomarker.md
biomarkers/vo2max-biomarker.md
biomarkers/grip-strength-biomarker.md
biomarkers/homa-ir-biomarker.md
biomarkers/lbp-biomarker.md

Parent protein/receptor pages (5 + 3 in flight):

molecules/proteins/tmsb4x.md (TB-500 parent — Thymosin β4)
molecules/proteins/bdnf.md (Semax/Dihexa-referenced BDNF)
molecules/proteins/ghsr.md (MK-677 ghrelin receptor)
molecules/proteins/fpr2.md (LL-37 immunomodulatory receptor)
molecules/proteins/alpha-msh.md (KPV/Melanotan-II parent peptide)
molecules/proteins/nefl.md (NfL biomarker parent)
molecules/proteins/mc1r.md (wave 3 — Melanotan II receptor)
molecules/proteins/mc4r.md (wave 3 — appetite/setmelanotide receptor)
molecules/proteins/mots-c-peptide.md (wave 3 — MOTS-c endogenous biology, complementing R36 compound page)

Major verifier findings — fabricated/inverted/wrong claims caught

This round was unusually high-yield for fabricated cohort numbers, mis-attributed cohorts, and direction-inverted claims. Every single seeded page had at least one substantive correction.

Fabricated cohort sample sizes (R28 lesson — seeder DOI/cohort memory unreliable):

NLR biomarker — three n fabrications: Fest 2018 Sci Rep (4,650 → 8,711), Welsh 2018 (393,368 → 478,259), Liu 2024 (387,739 → 448,574). None of the fabricated numbers appear in the source papers.
Adamstein 2021 (NLR): 10,061 was CANTOS alone; total across 5 trials = 60,087.
Disanto 2017 (NfL): wiki said n=219 MS + 40 controls; actual Lugano cohort = 142 MS + 254 HC; SMSC = 246 MS.
Niroula 2021 (CHIP): n was missing entirely (55,383 CHIP WES across UKB + MGBB).
Mattsson 2019 (NfL): cohort wrong (BioFINDER → ADNI); n wrong (1,182 → 1,583).

Direction-inverted claims (most consequential class):

Roberts 2020 LBP↔CVD risk (LBP biomarker) — wiki implied positive/harmful association; paper actually shows negative correlation (r = −0.462, p<0.0001). Higher Framingham CHD risk → LOWER LBP. Direction reversed.
Mattsson 2019 NfL trajectory in CU controls (NfL biomarker) — wiki said “increases in AD but not in cognitively normal controls”; paper actually shows NfL increases in ALL diagnostic groups (CU 2.4, MCI 2.7, AD 4.9 ng/L/yr; group differences p<0.001).
Picca 2022 ghrelin trajectory (GHSR receptor) — wiki said paper “identified ghrelin dysregulation as a candidate biomarker for anorexia of aging”; paper actually says plasma total/acyl-ghrelin “do not show variations with aging.” Reversed.
Wang 2014 SPM-receptor expression in AD (FPR2 receptor) — wiki said “elevated FPR2/LXA4R expression in AD brains (compensatory upregulation)”; paper Western blot shows protein NOT significantly different (z=−0.053, p=0.958); only ChemR23 (a different receptor) was elevated.
Wang 2014 hippocampal RvD1 in AD (FPR2) — wiki implied reduction; paper shows no significant difference (z=−0.575, p=0.62).
Fontana 2016 CR/IGF-1 (IGF-1 biomarker) — wiki implied CR lowers absolute IGF-1; paper shows null result (P=0.919); only molar ratio shifted.
Sayed 2021 iAge IL-6 role (IL-6 biomarker) — wiki said “IL-6 remains one of the strongest individual predictors within composite inflammatory aging models”; paper explicitly states canonical infection markers (IL-6, TNF-α) were “not major contributors to iAge” and IL-6 is absent from the top-15 Jacobians.

Major attribution errors:

Vitale 2012 (IGF-1 biomarker) — wiki labeled as “Ashkenazi longevity families”; paper is Italian multicenter (Milan/Bologna/Florence). Ashkenazi cohort is Milman 2014 (Barzilai).
Mandsager 2018 reference group (VO2max biomarker) — entire 5-row HR table was fabricated against “Below-Average” reference; paper actually compares vs Low as reference (Elite vs Low HR 0.20). Numbers 1.41/1.00/0.75/0.54/0.20 appear nowhere in the paper.
Tanaka 2000 citation (VO2max) — DOI/journal/pages cited (JACC 35:726-731) is a wrong paper entirely; that JACC paper is about heart rate not VO2max. Real Tanaka/Monahan/Seals 2001 is volume 37:153-156.
Singh 2013 DOI (LL-37, R36 verifier earlier) — wrong DOI resolved to a cyanophage paper.
Howard 1996 species comparison (GHSR) — wiki said “96% identity human/rat”; paper compares human/swine (~93% identical, ~98% similar); rat never mentioned.

Material non-significance hidden:

Adunsky 2011 CHF early termination (MK-677 R36 verifier) — trial was terminated early for CHF safety signal; missing from seeder. Authors concluded “unfavorable safety profile.”
Fujita 2023 cystatin C-adjusted GDF-15 — wiki implied persistent significance; actual CIs cross 1.0 (HR 1.65 [0.89-3.05]).
MAPT 2026 GDF-15 frailty incidence — wiki showed only minimally-adjusted Model 1 HR 1.69; fully-adjusted Model 4 HR is 1.40 (p=0.187, NS).
Erickson 2011 BDNF + memory (BDNF protein) — Time×Group interactions both NOT significant (memory p=0.40; BDNF p=0.23); seeder had implied between-group superiority. Within-aerobic-group BDNF↔volume correlation (r=0.36–0.37) is the actual finding.
Sosne 2015a (TMSB4X) — Phase 2 RCT FAILED both co-primary endpoints (P=0.26, P=0.22); the 27% reduction was a secondary endpoint.
Tan 2021 (TMSB4X cardiac) — Tβ4 ALONE did NOT improve LVEF; only Tb4+CM combination did.
Watts 2022 IGF-1/prostate MR (IGF-1 biomarker) — early-onset MR result OR 1.13 (95% CI 0.98–1.30, p=0.08) NOT significant; wiki had bare point estimates without CIs/p-values.

Fabricated outcomes / mechanism details (R33 lesson):

Dmitrieva 2010 neurotrophins (Semax, R36 verifier earlier) — added NT-4 + p75NTR not measured in paper.
NEFL phosphorylation sites — “Ser-2, Ser-55, Ser-57” wholly wrong; Ser-2 is N-acetylserine; correct sites are Tyr-43, Ser-56, Ser-67, Ser-103 (head/body) + Ser-472, Ser-502, Thr-520 (tail). Six IF-rod-domain coil/linker boundaries also wrong.
Niroula 2021 L-CHIP↔CAD/mortality (CHIP biomarker) — major clinical nuance missing: L-CHIP is NOT associated with CAD or all-cause mortality; only large M-CHIP confers CV risk.
Vlasschaert 2023 IL-6R-CHIP (CHIP biomarker) — protective effect specific to non-DNMT3A + large-VAF (≥10%) CHIP; missing from seeder.
β-MSH residue positions (alpha-MSH) — wrong (189–206 → 217–234 per UniProt P01189).
Dalmasso 2008 mouse n (alpha-MSH) — 4× error (40/group → 10/group).
Cooray 2013 SAA dimerization state (FPR2) — biased-agonism table had “Heterodimer-like”; SAA actively decreases ALX homodimer BRET signal; corrected.

Mis-cited methodological points:

HOMA-IR cutoffs (1.5/2.0/2.5) — NOT in either Matthews 1985 or Wallace 2004; tagged #gap/cutoffs-not-from-primary-source.
Wallace 2004 publication count — “500+ publications” (vague) → “572” (exact).
Kjaergaard 2025 MR (VO2max) — null result framing important causal caveat to strong observational evidence; preserved.

Cross-page propagation passes

interventions/lifestyle/exercise.md — Erickson 2011 framing corrected to remove between-group BDNF and memory overclaims; footnote summary updated to reflect non-significant Time×Group interactions.

Schema gaps surfaced (forward CLAUDE.md cleanup)

type: biomarker modality: enum lacks somatic-genetic for NGS-based assays (CHIP precedent). Used composite-clinical as placeholder.
type: biomarker mendelian-randomization: enum doesn’t distinguish receptor-axis vs direct-protein-pQTL MR (IL-6 / IL-6R precedent).
type: biomarker training-cohort: field is repurposed for single-analyte cytokines/biomarkers without ML training — convention now established, document in CLAUDE.md.

Forward queue surfaced

Implicit stubs flagged for future rounds:

[[il-6]] protein page (surprisingly absent — IL-6 has multiple cross-references but no dedicated protein page)
[[tocilizumab]], [[colchicine]] compound pages (cross-referenced from biomarker pages)
[[neurotrophin-signaling]], [[trkb-pathway]] pathway pages (BDNF cross-references)
[[spm-pathway]] pathway page (FPR2 cross-references)
[[POMC]] parent protein page (alpha-MSH parent; surface POMC processing biology)
[[MC3R]], [[MC5R]] (deferred from this round; lower aging-relevance)
[[melanocortin-system]] pathway/MOC page

Total agents dispatched (this round)

16 seeders (8 wave 1 + 5 wave 2 + 3 wave 3) + 19 verifiers (8 wave 1 + 8 wave 2 + 3 wave 3) = 35 invocations across the day. All 18 atomic content pages now verified: true with partial-scope notes.

Wave 3 highlights

MC1R (Q01726): afamelanotide FDA-approval (October 2019) verified; Gan-Or 2016 PD-meta-analysis framing corrected — paper does NOT condition on melanoma alleles; meta-analysis OR=1.22 (1.02–1.47, p=0.03) with significant heterogeneity (p=0.048); removing heterogeneous study gives OR=1.11 (p=0.27). Unsourced “sub-nanomolar Kd” claim removed. Feng 2015 reframed as review synthesis (was treated as primary).
MC4R (P32245): setmelanotide FDA structure corrected — LEPR was in original 2020-11-25 NDA along with POMC + PCSK1 (NOT a separate 2020 extension as wiki said); BBS added 2022-06-16 SUPPL-1. Haqq 2022 n corrected (44 BBS + 3 Alström → 32 BBS + 6 Alström = 38 total). Lotta 2019 n corrected (~550,000 → 452,300 UK Biobank). Collet 2017 mislabeled as “small case series” — actually Phase 1b randomised double-blind placebo-controlled RCT (n=6 setmelanotide + n=2 placebo + n=41 obese controls; mean weight loss −3.48 kg p<0.0001).
MOTS-c-peptide (A0A0C5B5G6): Zempo 2021 kinesio-genomic p-value corrected (p=0.01 → p=0.014) — wiki rounded to imply unmet significance threshold; cross-propagated back to the R36 mots-c compound page during this verifier pass. UniProt mutagenesis 13-16 vs 11-14 distinction clarified — 13-16 abolishes DNA binding but does NOT affect nuclear localization (only 11-14 affects both). D’Souza 2020 n=26/group added; ~11% middle-aged + ~21% older plasma MOTS-c decline magnitudes added.

[2026-05-08] ingest | Qi 2026 DHM-skin-aging epigenetic-clock paper (user-requested ad-hoc)

User-supplied URL: https://link.springer.com/article/10.1007/s13555-026-01764-4 — Qi M, Pitta P,… Falckenhayn C, Grönniger E (15 authors, Beiersdorf AG + DKFZ + Mahidol + UNIFESP + Howard) · Dermatology and Therapy 2026 · doi:10.1007/s13555-026-01764-4 · open-access CC BY-NC 4.0 · published online 2026-04-25.

Why ingested: new skin-specific epigenetic-clock paper directly relevant to the existing #gap/unsourced claim in skin-aging (epigenetic alterations subsection) and the epigenetic-alterations hallmark page. Adds first cosmetic-grade demonstration of methylation-clock movement at 8 weeks. Paper too recent for a local paper archive ingest; full PDF not yet retrieved.

Pages touched:

studies/qi-2026-dhm-epigenetic-skin-aging.md — NEW. Full study extraction from open-access HTML via WebFetch (two passes for verification consistency). Frontmatter: verified: false + auto-extraction banner. Captures: two-part design (n=17 multi-ethnic methylome pilot + n=60 Brazilian product-use single-cohort open-label, no vehicle control); new 23k epidermis clock (CV MAE 5.66 yr / multi-ethnic 4.88 yr / no Fitzpatrick bias); bivalent-region hypermethylation cross-ethnicity conservation; DHM-containing serum (DHM + low/high-MW HA + glycine saponin + enoxolone) twice-daily 8-wk + SPF 50+ → ~2.1 yr DNAm-age reduction (paired Wilcoxon p=0.029, 40% ≥5 yr responders, no CI), wrinkle −13.9% (p<0.001), echogenicity +10.4% (p<0.001), R=0.31 epigenetic age × wrinkle volume. Industry-funded (Beiersdorf); 8/15 authors are Beiersdorf employees, 3 more declare consultation fees.
phenotypes/skin-aging.md — UPDATED (verified=true preserved; this update adds new sourced content, does not modify previously verified claims). § “Epigenetic alterations — skin-specific epigenetic clocks” rewritten: added 4-row clock comparison table (Horvath 2013, Bormann 2016, Qi 23k 2026, TapeLift 2026); added bivalent-region-hypermethylation cross-ethnic-conservation finding; reframed the prior #gap/unsourced (UV-exposed vs unexposed clock acceleration) as #gap/needs-replication with explicit comment that Qi 2026 does not directly address paired anatomic-site comparison. § “Therapeutic landscape”: added new subsection “Topical DNMT-inhibitors — cosmetic-grade epigenetic-readout signal (Qi 2026)” before Interventional aesthetics. Five new footnotes: ^horvath2013, ^qi2026, ^bormann2016, ^rodriguezParedes2026 (last two flagged #gap/needs-pdf-verification).

Verification status: Qi 2026 study page is verified: false — auto-extracted from open-access Springer HTML (two WebFetch passes summarised by a fast model); full PDF not read end-to-end. Open-access CC BY-NC 4.0; the wiki-verifier subagent can flip the flag after PDF cross-check. Particular verification priorities: (1) Supp Fig S2 siRNA knockdown gene list, (2) precise statistical method for bivalent-region enrichment, (3) confirmation that n=17 pilot has zero overlap with n=60 product-use cohort, (4) full DOI/journal for Bormann 2016 + Falckenhayn 2024 + Rodríguez-Paredes 2026 references (all flagged #gap/needs-pdf-verification until confirmed).

Decision: NO compound page for DHM (dihydromyricetin / ampelopsin). This is a single industry-funded cosmetic intervention; mechanism-of-action evidence (DNMT inhibition) rests on Falckenhayn 2024 in vitro. Insufficient evidence to warrant a full compound page in the senotherapeutic / geroprotector tier; mention is contained within the Qi 2026 study page and the skin-aging therapeutic-landscape subsection. Reconsider if (a) vehicle-controlled replication appears, (b) DHM accumulates a literature beyond the cosmetic-skin niche, or (c) the wiki begins systematic seeding of cosmetic-grade interventions.

Decision: NO new biomarker pages for the 23k clock or TapeLift clock yet. Both are tissue-specific tape-strip clocks with single-paper provenance; the existing biomarker bay covers the major cross-tissue/blood clocks (Horvath 2013, Hannum 2013, PhenoAge 2018, GrimAge 2019, DunedinPACE 2022, Lehallier proteomic 2019). If a third independent skin-specific clock paper appears, reconsider seeding biomarkers/skin-methylation-clocks.md as an aggregate.

Forward-queue items:

Retry in 2–4 weeks (paper likely not yet in OpenAlex pipeline)
Once PDF is available: schedule wiki-verifier pass on Qi 2026 study page; cross-check Bormann 2016 + Falckenhayn 2024 + Rodríguez-Paredes 2026 reference DOIs
Optional: brief abstract-level lookup of Falckenhayn 2024 (DHM as DNMT inhibitor in vitro) to check whether it warrants a separate study page anchoring the mechanism claim

[2026-05-09] verify — molecules/proteins/tmsb4x.md

Pages verified: 1 (partial scope)

Sources verified against full PDF:

Tan 2021 (Theranostics, doi:10.7150/thno.56757) — downloaded and confirmed end-to-end
Sosne 2015 OPTH (Clinical Ophthalmology, doi:10.2147/OPTH.S80954) — downloaded and confirmed end-to-end
Nguyen 2025 (IOVS, doi:10.1167/iovs.66.14.31) — downloaded and confirmed (first 4 pages)

Sources verified against abstracts only (not_oa):

Philp 2004 (MAD, doi:10.1016/j.mad.2003.11.005) — PMID 15037013 abstract confirmed
Sosne 2010 (FASEB J, doi:10.1096/fj.09-142307) — abstract confirmed via Crossref

Sources unverifiable (not_oa):

Malinda 1997, Goldstein 2015, Bjorklund 2020, Bock-Marquette 2023, Sosne 2015 Cornea — all not_oa
Badamchian 2007 — not_oa; no-fulltext-access tag added to footnote

Canonical IDs verified:

UniProt P62328, NCBI Gene 7114, HGNC 11881, Ensembl ENSG00000205542 — all confirmed via REST APIs
NCT05555589 Phase 3 RECRUITING — confirmed via ClinicalTrials.gov v2 API

Sequence length RESOLVED:

44 aa = canonical sequence including Met-1 (UniProt P62328)
43 aa = mature processed chain (Met-1 removed, per UniProt “Initiator methionine: Removed” annotation)
Literature’s “43 aa” convention refers to the mature form — both counts are correct in their contexts
LKKTETQ (TB-500 fragment) = positions 18–24 in canonical 44-aa numbering, or 17–23 in mature-chain numbering
Sequence note in page updated to document this resolution

Corrections made:

Sequence note: “44 aa canonical, 43 aa convention unexplained” → explains Met-1 cleavage resolves discrepancy; mature processed chain = 43 aa
Sosne 2015a RCT (MAJOR): “significant corneal/conjunctival staining improvement vs placebo; no adverse events” → corrected to: primary endpoints FAILED (inferior corneal staining P=0.2586, ocular discomfort P=0.2210); 27% discomfort reduction was a secondary endpoint at day 28 CAE (P=0.0244); significant improvements in central (P=0.0075) and superior staining (P=0.0210) only; inferior and conjunctival differences NS; 2 mild ocular TEAEs in Tβ4 group (none drug-related)
Tan 2021 cardiac (MAJOR): “improved left ventricular systolic function” attributed to Tβ4 microspheres → corrected to: only the Tb4+CM combination significantly improved LVEF; Tb4 alone reduced scar area but did not significantly improve LVEF vs MI controls; arrhythmia had overall Fisher p=0.038 (not “no arrhythmia”)
Tan 2021 in-vitro dose context: added note that 600 ng/mL is a supraphysiological in-vitro concentration; dose-response-unclear added
Sosne 2010 footnote: WH2 domain “aa 17-24” → “aa 17–23 in mature-chain numbering” (corrected off-by-one)
Nguyen 2025 footnote: clarified paper calls NCT05555589 “phase 2” while ClinicalTrials.gov lists it as Phase 3
Philp 2004 footnote: added PMID 15037013, confirmed abstract language, tagged no-fulltext-access
Badamchian 2007 footnote: archive status “not confirmed” → “confirmed (not_oa)”; added no-fulltext-access
PTM section: corrected “Lys-12 (alternate)” annotation; added SUMO2 isopeptide cross-link clarification
Phase 3 trial entry: added sponsor (ReGenTree LLC), n=70, completion 2026-05-30, Nguyen 2025 discrepancy note

Downstream pages that may need updates:

tb-500 — may inherit the overstatement that “Tβ4 alone improves LV function”; should reflect combination-only result from Tan 2021
tb-500 — WH2 domain position reference should use mature-chain 17–23 consistently
Any intervention/pharmacological or hallmarks page citing the Sosne 2015a result as “primary-endpoint success”

[2026-05-09] verify — biomarkers/gdf-15-biomarker.md

Pages verified: 1 (partial scope — 4 sources PDF-verified; 2 sources abstract-verified; 2 sources not in archive)

Sources verified against full PDF:

Webber 2024 (medRxiv preprint; PMC11326340) — PDF downloaded and verified
Chen 2026/Protein & Cell (peer-reviewed) — PDF downloaded and verified
Hansen 2026/GeroScience (peer-reviewed) — PDF downloaded and verified
MAPT 2026 / Sánchez-Sánchez 2026 / JCSM (peer-reviewed) — PDF downloaded and verified

Sources verified via PubMed abstract only (closed-access):

Wiklund 2010 (Aging Cell) — bronze OA 403; n and OR confirmed from abstract (PMID corrected: 20726902 → 20854422)
Kim 2022 (J Gerontol A) — closed-access; values confirmed from abstract
Fujita 2023 (J Gerontol A) — closed-access; values confirmed from abstract

Sources not in archive / not checked:

Kempf 2007 (Clinical Chemistry) — download failed; existing footnote note retained; values not independently confirmed
Dallmeier 2016 (Clinical Chemistry) — DOI not found in archive; values not independently confirmed

Corrections made:

Wiklund 2010 PMID: 20726902 (wrong — resolved to a Swiss HIV-pregnancy paper) → 20854422 (confirmed correct). Values unchanged (OR 3.38 confirmed).
Kim 2022 cohort label: “Korean BOSA cohort” → “Korean Frailty and Aging Cohort Study” (correct per abstract and Crossref). Fixed in body, footnote, and frontmatter training-cohort.
Fujita 2023 HR attenuation: wiki stated “attenuated to 1.65 after adjusting for cystatin C” without CI, implying maintained significance → corrected to show both cystatin C-adjusted (HR 1.65, 95% CI 0.89–3.05, NS) and β2-microglobulin-adjusted (HR 1.69, 95% CI 0.91–3.12, NS) results are non-significant. This is a material correction: the wiki previously implied a significant attenuated association; it is actually non-significant when renal function is adjusted.
MAPT 2026 mean age: 75.2 → 75.3±4.4 (from Table 1: 75.30±4.37)
MAPT 2026 frailty tool: now explicitly named as Fried phenotype ≥3 criteria (not stated previously)
MAPT 2026 incident frailty HR: added critical nuance that HR=1.69 (1.03–2.78) is from minimally adjusted Model 1 (age+sex only) and becomes non-significant in fully adjusted Model 4 (HR=1.40, 0.85–2.33, p=0.187). This materially weakens the prospective GDF-15 → frailty claim.
Hansen 2026 AUC CI: AUC 0.681 → 0.681 (95% CI 0.623–0.739); sarcopenia AUC 0.577 added (poor discriminatory ability — newly added detail)
Hansen 2026 frailty instrument: Clinical Frailty Scale ≥5 (newly stated)
Chen 2026: “14-year follow-up” → “median 14.5-year follow-up (IQR 13.8–15.3)”; mediation proportion 25.7% for GDF-15 added; clarified paper is peer-reviewed (Protein & Cell), not a preprint
Webber 2024: confirmed no peer-reviewed version as of 2026-05-09; added PMC11326340; full author list added; “verified against preprint PDF” note added
Hansen 2026 journal year: “GeroScience 2026” → “GeroScience 2025, 48:1955–1966” (published November 2025)

Unverifiable claims remaining:

Senolytics → GDF-15 reduction: no published human RCT found; unsourced tag retained
Kempf 2007 median 762 pg/mL: footnote notes pending status; value not independently confirmed
Dallmeier 2016 HR values: DOI not in archive; values not confirmed

Downstream pages that may need updates:

gdf15 — may carry Fujita 2023 claims; ensure non-significance after renal adjustment is noted
cellular-senescence — if it cites senolytics → GDF-15 reduction as evidence; that claim remains unsourced
senolytics — same caveat

[2026-05-09] verify — molecules/proteins/nefl.md

Pages verified: 1 (partial scope — 3 of 4 primary sources verified against PDF; Lariviere 2004 not_oa)

Sources verified against PDF:

Mersiyanova 2000 Am J Hum Genet (doi:10.1086/302962) — downloaded and read in full (10 pp)
Yuan 2017 Cold Spring Harb Perspect Biol (doi:10.1101/cshperspect.a018309) — downloaded and read in full (25 pp)
Khalil 2020 Nat Commun (doi:10.1038/s41467-020-14612-6) — local PDF; read in full (9 pp)

UniProt P07196 verified via REST API (rest.uniprot.org).

Sources unverifiable:

Lariviere 2004 J Neurobiol (doi:10.1002/neu.10270) — not_oa; load-bearing claims re-anchored to Yuan 2017 and Mersiyanova 2000; tagged no-fulltext-access

Corrections made:

Domain residue table (MAJOR): 6 of 8 coil/linker boundary values wrong vs UniProt P07196. Corrected: Coil 1A 90–153 → 93–124; Linker L1 154–166 → 125–137; Coil 1B 167–248 → 138–234; Linker L12 249–289 → 235–252; Coil 2A 290–316 → 253–271; Coil 2B 317–400 → 281–396. Tail A/B boundaries (397–443; 444–543) confirmed correct.
PTM phosphorylation sites (MAJOR): “Ser-2, Ser-55, Ser-57” → actual UniProt P07196 sites: Tyr-43, Ser-56, Ser-67, Ser-103 (head-domain/body); Ser-472, Ser-502, Thr-520 (tail). Ser-2 is an N-acetylserine (initiator Met removed), not a phosphorylation site.
O-GlcNAc glycosylation sites: “Serine/threonine residues” (generic) → “Thr-21 and Ser-27” (UniProt P07196)
NF-H predicted MW: 112.4 → 112.5 kDa (Yuan 2017: “112.5, 102.5, 61.5 kDa” in human)
NF-H KSP repeat count: “up to ~50 sites” → “45 KSP repeats in human, 51 in mouse” (Yuan 2017 Section 5)
Mersiyanova 2000 footnote n (MAJOR): “n=1 Russian family (pedigree + 180 controls)” → “n=28 family members (six-generation; 12 affected, 14 unaffected, 2 unknown) + 46 unrelated CMT patients + 90 healthy controls (180 chromosomes)”. The “180 chromosomes” figure was not total controls; it was the chromosome count from 90 individuals screened specifically for the Gln333Pro variant.
Khalil 2020 longitudinal n: “n=103 longitudinal” → “n=103 agreed to follow-up; n=95 final longitudinal dataset” (8 excluded for incident disease during follow-up)
Khalil 2020 cohort name: “Graz Longitudinal Study” → “Austrian Stroke Prevention Study / ASPS-Fam (Graz, Austria)”
Khalil 2020 body text: “accelerates sharply after age 60” → accurate statement: stable below 60, nonlinear increase above 60 with significant variance increase (Brown-Forsythe p<0.0001); Spearman rs values added (cross-sectional brain volume −0.321; longitudinal brain atrophy −0.290, p<0.01)
Footnote archive statuses: Mersiyanova 2000 and Yuan 2017 updated to “downloaded”
Auto-extraction banner removed

Unverifiable claims remaining:

Slow axonal transport velocity “~0.2–1 mm/day (SCa)” — not source-verified in this pass; Yuan 2017 discusses slow transport qualitatively without giving exact velocity values in sections read
Lariviere 2004 scaffold/ALS/aggregate claims — not_oa; re-anchored to Yuan 2017 + Mersiyanova 2000

Downstream pages that may need updates:

nfl-biomarker — cites Khalil 2020; the longitudinal n (95) and cohort name corrections were also applied in the earlier nfl-biomarker verify pass today; confirm consistent

[2026-05-09] verify — interventions/dietary/mediterranean-diet.md

Pages verified: 1

interventions/dietary/mediterranean-diet.md — corrections: 9 (see below); verified: true (partial scope)

Sources checked:

PREDIMED 2018 (doi:10.1056/NEJMoa1800389) — PDF downloaded and verified (full read)
CORDIOPREV 2022 (doi:10.1016/S0140-6736(22)00122-2) — not_oa; verified against PubMed abstract (PMID 35525255)
Lyon Diet Heart 1999 (doi:10.1161/01.cir.99.6.779) — not_oa; verified against PubMed abstract (PMID 9989963)
Bozack 2025 (doi:10.1038/s41514-025-00298-x) — PDF downloaded and verified (full read)
Nucci 2026 (doi:10.1016/j.nut.2026.113189) — verified against PubMed abstract (PMID 41930792)
Tran 2024 (doi:10.1017/S0007114524001648) — verified via Crossref (title, journal, first author); full text not read
García-Calzón 2016 (doi:10.1016/j.clnu.2016.03.013) — not verified; PDF not downloaded; claims correctly tagged observational

Corrections:

PREDIMED site count: “57 centres” → “11 sites encompassing 169 clinics” (per paper methods)
PREDIMED age range: “55–80 adults” → “men 55–80 / women 60–80” (per enrollment criteria)
PREDIMED HRs with CIs: added exact ITT adjusted HR 0.69 (0.53–0.91) EVOO; 0.72 (0.54–0.95) nuts (previously “~0.70 / ~0.72” with no CIs)
PREDIMED primary endpoint: confirmed as MI + stroke + CV death (stroke includes fatal, not “non-fatal stroke” as implicit from prior wording); footnote updated to reflect per-arm n breakdown
CORDIOPREV HR: “~0.74 (verify CI)” → HR range 0.719 (0.541–0.957) to 0.753 (0.568–0.998); events 87 vs 111; p=0.039; sex-stratified finding added (benefit in men only)
Lyon risk ratio: “~0.27–0.30 for cardiac death + non-fatal MI combined” → adjusted RR 0.28–0.53 across 3 pre-specified composite outcomes; event counts added (CO1: 14 vs 44; CO2: 27 vs 90; CO3: 95 vs 180)
Bozack 2025 PhenoAge coefficient: −1.02 yr → −1.05 yr (95% CI −1.87, −0.22); clock count 5→11; DunedinPACE→DunedinPoAm; MiAge removed (not in paper)
Kasiri 2024 → Tran 2024 first-author attribution; footnote tag renamed [^kasiri2024]→[
Nucci 2026: DOI, PMID, and exact pooled RR 0.96 (0.95–0.97) added; footnote seeded
clinical-trials-active: 4 → 325 (per ClinicalTrials.gov v2 RECRUITING+ACTIVE_NOT_RECRUITING countTotal query 2026-05-09)

Unverifiable claims remaining:

García-Calzón 2016 telomere substudy — PDF not downloaded; body text correctly flags as observational/prospective; magnitude and significance not extracted
Tran 2024 CRP umbrella review — exact effect sizes and certainty grading not extracted from full text
Oleocanthal dose-equivalence (Beauchamp 2005 Nature letter) — not re-verified; correctly tagged dose-response-unclear

Supersession check (R25):

Recent meta-analyses (2023–2026) reviewed: Nucci 2026 (all-cause mortality, RR 0.96/point adherence, n=1.8M), Brunello/Nucci 2026 (oncology). No paper contradicts PREDIMED/CORDIOPREV core findings. CORDIOPREV (2022) and Bozack 2025 are the most recent primary sources; neither superseded.
literature-checked-through: 2026-05-09

Downstream pages that may need updates:

None yet identified; this is the first mediterranean-diet page; no entity pages currently cite it via study footnotes

[2026-05-09] verify — tocilizumab.md

Pages verified: 1

molecules/compounds/tocilizumab.md — corrections: 7 (see below); verified: true (partial scope)

Sources checked:

AMBITION Jones 2010 (doi:10.1136/ard.2008.105197) — PDF locally available; fully read; no corrections needed
RECOVERY 2021 (doi:10.1016/S0140-6736(21)00676-0) — PDF downloaded and verified
GiACTA Stone 2017 (doi:10.1056/NEJMoa1613849) — PDF downloaded and verified
Cupido 2022 (doi:10.1111/bcp.15191) — PDF downloaded and verified
OPTION Smolen 2008 (doi:10.1016/S0140-6736(08)60453-5) — not_oa; not verified; gap tagged
TOWARD Genovese 2008 (doi:10.1002/art.23940) — not_oa; not verified; gap tagged
Biggioggero 2018 review and Lu 2023 — not downloaded; not verified

Corrections:

RECOVERY p-value: “P=0.003” → “p=0.0028”
RECOVERY n breakdown: added per-arm (2022 tocilizumab, 2094 usual care)
RECOVERY corticosteroid claim: “consistent regardless of corticosteroid co-use” qualified; benefit larger in users (RR 0.79) vs non-users (RR 1.16); authors note interaction may be chance
RECOVERY discharge endpoint: added rate ratio 1.22 (95% CI 1.12–1.33)
GiACTA biweekly arm remission: “53.7%” → “53%”
GiACTA design: 4-arm trial (weekly/biweekly/placebo+26wk/placebo+52wk); not “all with 26-week taper”
Cupido 2022: IL6R instrument is polygenic (up to 19 variants), not primarily Asp358Ala; added omitted pneumonia risk signal (OR 1.17, 95% CI 1.09–1.27)

Unverifiable claims remaining:

OPTION n=623 and ACR20 figures — not_oa
TOWARD n=1,220 and ACR20 figures — not_oa
Vargas-Maciel frailty pilot — unsourced
ChEMBL/DrugBank IDs not re-checked against live databases

Downstream pages that may need updates:

None identified; tocilizumab is the canonical compound page; no downstream entity pages currently cite it via study footnotes

[2026-05-09] verify — molecules/proteins/hmgcr.md

Page: molecules/proteins/hmgcr.md · type: protein · verified: false → true

Sources checked: UniProt P04035 (REST API + FASTA sequence); CTT 2010 Lancet (local PDF); Swerdlow 2015 Lancet (local PDF); Horton 2003 PNAS (local PDF); Sever 2003 Mol Cell (local PDF); Benn 2017 BMJ (local PDF); CTT 2024 Lancet Diabetes Endocrinol (local PDF); Ference 2015 JACC (PMC full text; local PDF failed download).

Corrections made (7):

Active-site residues — critical: Asp559 → Glu559, Asp691 → Lys691 (seeder had correct residues from training memory; the wiki page then incorrectly “corrected” them to Asp. UniProt sequence P04035 position 559=E, 691=K confirmed.)
T2D risk %: “9–11% (RR)” → 12% (OR 1.12, 95% CI 1.06–1.18) per Swerdlow 2015 RCT meta-analysis; dose-intensity breakdown added from CTT 2024.
Ference 2015 n: “~48,000” → 108,376 (primary analysis; external validation up to 189,539); effect size detail added.
Benn 2017 n: “~75,000” → 111,194 (Copenhagen General Population Study + City Heart Study combined).
Benn 2017 PCSK9/AD claim: removed incorrect statement that “PCSK9-lowering alleles show AD signal.” Both HMGCR and PCSK9 alleles were NS for AD in MR (causal RR 0.57, 95% CI 0.27–1.17 for HMGCR; not significantly different for PCSK9).
Sever 2003 cell model: “HEK293 + CHO” → primary model CHO-K1 (endogenous studies used HEK-293S for Fig. 4 only).
Aging-relevance T2D tradeoff text: “~9% RRR” → “~10–12% standard intensities, up to ~36% high intensity.”

CTT 2024 integration: CTT 2024 Lancet Diabetes Endocrinol adds dose-intensity T2D breakdown. Framing consistent with Swerdlow 2015; no supersession of cardiovascular efficacy evidence.

literature-checked-through: 2026-05-09 added to frontmatter.

Downstream pages that may need updates:

interventions/pharmacological/statins.md (if it exists) — T2D risk figures and Benn 2017 AD characterization
molecules/proteins/pcsk9.md — if it describes a PCSK9-specific AD-protective signal, cross-check against Benn 2017 (both targets were NS in MR)

[2026-05-09] update — molecules/compounds/taurine.md (R35 user-supplied additions)

Page: molecules/compounds/taurine.md · type: compound · verified: true (already true; updated verified-scope and literature-checked-through)

User context: user supplied two additional 2026 Frontiers taurine review URLs to incorporate (one day after R34 backfill).

Sources added (both 2026 narrative reviews; landing-page full text fetched via WebFetch; not yet PDF-verified end-to-end):

Chen & Niu 2026 — Frontiers in Physiology 17:1809107; doi:10.3389/fphys.2026.1809107; gold OA. Narrative review of taurine + glutamine + exercise interactions in aging neuro-skeletal-muscle decline; aggregates ~7 taurine RCTs and ~9 glutamine RCTs across elderly women, healthy elderly men, heart failure, postmenopausal women, T2D, sarcopenic obesity, and an 8-yr Japanese cohort (n=1,454).
Zhang 2026 — Frontiers in Nutrition; doi:10.3389/fnut.2026.1783074; gold OA. Narrative review of taurine + exercise across obesity, aging, and diabetes; aggregates ~25 preclinical and clinical studies; comprehensive mechanistic framework covering adipose plasticity, metabolic flexibility, neuroinflammaging, PI3K/Akt cardiac remodelling.

Mechanistic additions to the page (not previously surfaced):

Adipose-tissue browning — taurine + exercise upregulates CIDEA, PGC-1α, PRDM16, UCP1/UCP2 + FAO genes (CPT1, PPARα/γ, LPL, HSL, ACOX1, CD36) in subcutaneous WAT in obese-women cohorts; adipocyte area decreases, RMR increases 151–335 kcal in 16-week trials, plasma irisin elevated post-bout in taurine arm only.
PI3K/Akt cardiac remodelling — combined taurine + endurance/resistance exercise normalises Bax/Bcl2 ratio (p≤0.001) and reduces caspase-3/-9 in STZ-diabetic Wistar rat hearts; PI3K gene expression rises ~57%; collagen deposition decreases; renin-angiotensin axis activity decreases. All preclinical / rat-only.
BDNF-TrkB neuroplasticity — exercise-dependent: MMSE preservation in elderly women only in combined arm, not taurine-alone.

Frontmatter updates:

literature-checked-through: 2026-05-08 → 2026-05-09
verified-scope: extended with R35 note documenting Chen & Niu 2026 + Zhang 2026 landing-page-text verification; full-PDF verification deferred

Footnotes added: [^chen2026], [

Net effect on page-level conclusions: None. The Singh-2023 vs Fernandez-2025 biomarker dispute is unchanged; no Yadav-lab or independent Singh-2023 lifespan replication identified in either review; Sharma 2025 mTOR/leukaemogenesis concern is not addressed by either review. Aging-intervention status remains “contested and unresolved.” The cardiometabolic + adipose-browning evidence base is now slightly broader, with stronger emphasis on taurine + exercise pairing.

Limitations of these additions:

Both papers are narrative reviews (not meta-analyses) — no new pooled effect sizes
Most browning-gene readouts trace to a single Brazilian research lineage (Batitucci/De Carvalho)
Cardiac-remodelling apoptosis-pathway evidence is rat-only
Neither review addresses the Singh-2023 lifespan-extension claim or its dependency on a deficit-replacement narrative
Landing-page text only; PDF verifier follow-up recommended for full quantitative cross-check

Downstream pages that may need updates:

molecules/compounds/glutamine.md — does not yet exist; Chen & Niu 2026 covers ~9 glutamine RCTs in aging context relevant to a future glutamine seed
phenotypes/sarcopenia.md — taurine + glutamine + exercise framing relevant if not already covered
No changes to study pages (pattern: 2026-05-08 backfill additions also footnote-only with no dedicated study pages)

🧬 Aging Wiki

Explorer

R35

log/R35.md — Round 35 entries

[2026-05-09] verify — molecules/proteins/growth-hormone-receptor.md

[2026-05-09] verify — molecules/proteins/c-met.md

[2026-05-09] verify — molecules/proteins/vegfr2.md

[2026-05-09] verify — molecules/proteins/enos.md

[2026-05-09] verify — biomarkers/glycanage-2017.md

[2026-05-09] verify — molecules/proteins/mc3r.md

[2026-05-09] verify — pathways/camp-signaling.md

[2026-05-09] verify — pathways/pomc-processing.md

[2026-05-09] verify — molecules/proteins/hgf.md

[2026-05-09] verify — molecules/compounds/colchicine.md

[2026-05-09] verify — molecules/proteins/mc5r.md

[2026-05-09] verify — processes/melanogenesis.md

[2026-05-09] verify — pathways/trkb-pathway.md

[2026-05-09] verify — pathways/spm-pathway.md

[2026-05-09] verify — pathways/neurotrophin-signaling.md

[2026-05-09] verify — molecules/proteins/pomc.md

[2026-05-09] verify — molecules/proteins/il-6.md

[2026-05-09] verify — melanocortin-system.md (MOC)

[2026-05-09] verify — mc4r.md

[2026-05-09] verify — mc1r.md

[2026-05-09] verify — setmelanotide.md

[2026-05-09] verify — mots-c-peptide.md

[2026-05-09] R35-Stage2 + R36-followup close — biomarkers + parent proteins/receptors

Pages added (wave 1 + 2; wave 3 still in flight)

Major verifier findings — fabricated/inverted/wrong claims caught

Cross-page propagation passes

Schema gaps surfaced (forward CLAUDE.md cleanup)

Forward queue surfaced

Total agents dispatched (this round)

Wave 3 highlights

[2026-05-08] ingest | Qi 2026 DHM-skin-aging epigenetic-clock paper (user-requested ad-hoc)

[2026-05-09] verify — molecules/proteins/tmsb4x.md

[2026-05-09] verify — biomarkers/gdf-15-biomarker.md

[2026-05-09] verify — molecules/proteins/nefl.md

[2026-05-09] verify — interventions/dietary/mediterranean-diet.md

[2026-05-09] verify — tocilizumab.md

[2026-05-09] verify — molecules/proteins/hmgcr.md

[2026-05-09] update — molecules/compounds/taurine.md (R35 user-supplied additions)

Graph View

Table of Contents

Backlinks