R20

log/R20.md — Round 20 entries

Sub-file of log — see parent for index.

[2026-05-05] verify | cbp-p300

molecules/proteins/cbp-p300.md

• page: molecules/proteins/cbp-p300.md
• sources verified against full PDF (6):
  • Gu & Roeder 1997 (10.1016/s0092-8674(00)80521-5, Cell 90:595–606) — 12 pp read end-to-end; p53 K-site acetylation founding paper
  • Petrij 1995 (10.1038/376348a0, Nature 376:348–351) — 4 pp read end-to-end; RSTS1 cohort and CREBBP locus paper
  • Roelfsema 2005 (10.1086/428172, AJHG 76:572–580) — 9 pp read end-to-end; RSTS2 EP300 cohort paper
  • Lee & Finkel 2009 (10.1074/jbc.M804232200, JBC 284:6322–6328) — 7 pp read end-to-end; autophagy substrate acetylation paper
  • Rodgers 2005 (10.1038/nature03354, Nature 434:113–118) — 6 pp read end-to-end; PGC-1α/SIRT1 acetylation axis
  • Heintzman 2009 (10.1038/nature07829, Nature 459:108–112) — 5 pp read end-to-end; H3K27ac/p300 enhancer ChIP paper
• sources unverifiable (not_oa): Thompson 2004 (10.1073/pnas.0408285101), Brunet 2004 (10.1126/science.1094637) — tagged no-fulltext-access
• sources not yet downloaded (OA pending): Karanam 2006 (10.1021/bi0522520), Lasko 2017 (10.1038/nchembio.2516), Chen 2002 (10.1126/science.1074918) — tagged no-fulltext-access where load-bearing
• domain coordinates verified against UniProt Q92793 (CREBBP) and Q09472 (EP300) via REST API
• corrections made (10):
  1. p53 K-site list: K386 removed — Gu 1997 explicitly states “K386 appeared completely unacetylated in comparison with the background”; list corrected to K370/K372/K373/K381/K382
  2. p53 mechanism: MDM2-displacement mechanism (K382ac competing with ubiquitination) removed from the p53 section — not in Gu 1997; attributed to “later work not directly demonstrated in Gu 1997”
  3. p53 conformational activation: fold-change quantified — “~20- to 30-fold over unmodified p53” added from paper (was unquantified in wiki)
  4. PGC-1α: 13 specific K-sites listed (K77/K144/K183/K253/K270/K277/K320/K346/K412/K441/K450/K757/K778) — wiki had “multiple acetylated lysines” without numbering
  5. PGC-1α writer attribution: explicit caveat added — “Writer identity not established by Rodgers 2005 — p300/CBP as acetyltransferase is inferred from the family’s known substrate scope, not demonstrated in this paper” — added in substrate table row, body section, AND footnote
  6. PGC-1α body: quantitative repression (nicotinamide-treated cells with hyperacetylated PGC-1α showed 24-fold repression) added; mouse age/strain (4-week-old C57Bl/6) and condition (fasted 24 h) added
  7. Petrij 1995 footnote n: “n=24” → “n=16 RSTS patients examined by protein truncation test (PTT); truncated proteins found in 2 patients” — paper never states 24; the PTT experiment is described as 16 patients
  8. Heintzman 2009 footnote: “9 cell types” → “5 human cell lines (HeLa, GM06690, K562, ES, BMP4-induced ES)” — confirmed from paper Methods
  9. Non-OA footnotes: Thompson 2004, Brunet 2004 tagged no-fulltext-access
  10. Pending-OA footnotes: Karanam 2006, Lasko 2017, Chen 2002 tagged no-fulltext-access where load-bearing
• incidental finding (not corrected on this pass): ep300.md domain coordinates for ZZ (listed ~1723–1770; UniProt Q09472 shows 1665–1713) and TAZ2 (listed ~1834–1890; UniProt shows 1728–1809) appear incorrect — flagged for main agent propagation
• verified flag: false → true (partial scope — see verified-scope field); ⚠️ banner removed
• downstream pages to check (main agent):
  • molecules/proteins/p53.md — may include K386 in p53 acetylation list attributed to Gu 1997; also check for unsourced MDM2-displacement claim
  • pathways/p53-pathway.md — may inherit MDM2-displacement mechanism attributed to Gu 1997
  • molecules/proteins/pgc-1alpha.md — may attribute CBP/p300 as the acetyltransferase writer citing Rodgers 2005 (unsupported)
  • molecules/proteins/sirt1.md — same PGC-1α writer attribution issue
  • molecules/proteins/ep300.md — domain coordinate discrepancies for ZZ (1723–1770 → 1665–1713) and TAZ2 (1834–1890 → 1728–1809) per UniProt Q09472 (separate verification task needed)
• Resolve Lu 2020 vision-restoration mouse-age minor discrepancy: SOX2 verifier said “3- and 12-month-old” for aging vision restoration; KLF4 verifier said “3- and 11-month-old”. Pages now consistent on the high-level point (NOT 5/20-mo). Re-read PDF on next lu-2020-osk-vision-restoration study-page seeding.
• Closed-access papers needing institutional access to verify: Matsumura 2016 (Science aad4395), Nishimura 2005 (Science), Wood 2009 (FASEB J), Whiting 1993 (JAAD), Premanand 2018, Scaffidi & Misteli 2006 (Science; PDF download failed), Sforza 2009 (Forensic Sci Int), Verzijl 2002 (Arthritis Rheum), Ferri 2004 (Development), Katz 2002 (Development), Segre 1999, Devries 2014 (MSSE), Avgerinos 2018 (Exp Gerontol), Dalla-Favera 1982.

[2026-05-05] verify | molecules/proteins/drp1.md

• page: molecules/proteins/drp1.md
• sources verified against full PDF (5): Bordt 2017 (devcel), Wakabayashi 2009 (JCB), Cribbs-Strack 2007 (EMBO Rep), Chang-Blackstone 2007 (JBC), Twig 2008 (EMBO J)
• sources unverifiable — not_oa (4): Smirnova 2001 (MBC), Otera 2010 (JCB), Friedman 2011 (Science), Waterham 2007 (NEJM)
• sources unverifiable — OA download failed (1): Cho 2009 (Science, 10.1126/science.1171091) — green OA status but 0 candidate URLs
• UniProt O00429 verified against live REST API: domain boundaries, PTM sites, isoform count, MW all confirmed
• corrections made (6):
  1. En1-Cre, not Nestin-Cre — brain-specific KO (Wakabayashi 2009) used engrailed-1 (En1-Cre), not Nestin-Cre. This is a material error affecting the phenotype description.
  2. Death at 36 h, not 3 weeks — En1-DrpKO mice die within 36 h of birth (neonatal lethality), not “3 weeks postnatally.” Phenotype is cerebellar developmental defect (hypoplasia, Purkinje cell loss), not progressive neurodegeneration.
  3. Mdivi-1 DRP1 Ki > 1.2 mM, not IC50 ~1–10 µM — Bordt 2017 shows human DRP1 GTPase Ki > 1.2 mM. The “IC50 ~1–10 µM” claim was for yeast Dnm1, not human DRP1. Removed from body; correct Ki added.
  4. Mdivi-1 effective concentration range: 25–100 µM, not 10–50 µM — Bordt 2017 demonstrates Complex I effects at 25 µM and above (not starting at 10 µM).
  5. Ser616 CDK1 citation attribution corrected — Chang & Blackstone 2007 covers only Ser637/PKA, not Ser616. PTM table Ser616 row now cites UniProt + PubMed IDs; Chang-Blackstone footnote restricted to Ser637.
  6. Cribbs-Strack 2007 uses rat Ser656, not human Ser637 — added explicit isoform-numbering note; rat Ser656 = human Ser637 (confirmed from paper and UniProt).
• added no-fulltext-access tags to: Smirnova 2001 claims, Otera 2010 claims, Friedman 2011 claims, Waterham 2007 claims, Cho 2009 claims
• gaps remaining unsourced: CDK5/Ser616 in AD neurons, alpha-syn/DRP1 direct interaction, age-dependent pSer616/pSer637 ratio in mouse brain
• downstream pages to check: alzheimers-disease.md, parkinsons-disease.md, mitophagy.md, mitofusins.md — may carry Mdivi-1 specificity claims without the Bordt 2017 Complex I caveat; also check for Nestin-Cre / En1-Cre confusion

[2026-05-05] verify | pathways/mitochondrial-dynamics.md

• page: pathways/mitochondrial-dynamics.md
• type: synthesis pathway page (lighter verification — claims inherited from atomic pages)
• sources verified against full PDF (2 newly downloaded):
  • Frank 2001 (10.1016/s1534-5807(01)00055-7, Dev Cell 1:515–525) — 11 pp read end-to-end
  • Karbowski 2002 (10.1083/jcb.200209124, JCB 159(6):931–938) — 8 pp read end-to-end
• sources already verified on atomic pages (4):
  • Twig 2008 — verified on drp1.md R12
  • Bordt 2017 — verified on drp1.md R12
  • Frezza 2006 — verified on opa1.md R10d and cytochrome-c.md
  • Otera 2010 / Friedman 2011 — remain not_oa; carry no-fulltext-access on drp1.md
• cross-link integrity: all 6 implicit stubs (mff, fis1, mid49, mid51, oma1, yme1l) confirmed slug-consistent with drp1.md and opa1.md; no slug drift; 10 remaining wikilinks all resolve to existing files
• provenance audit: Mdivi-1 Ki >1.2 mM, 80–85% cristae cytochrome c, DRP1-K38A apoptosis delay, Twig 2008 fission-then-mitophagy model, disease associations — all correctly attributed to their atomic-page sources
• corrections made (2):
  1. Frank 2001 apoptosis-delay quantified: “significantly delays caspase activation and cell death” → “significantly delays mitochondrial cytochrome c release and apoptotic cell death: 77% Drp1_wt COS-7 cells TUNEL+ at 10h vs 19% K38A; anti-Drp1 microinjection in HeLa reduces apoptosis from 86% to 17% at 20h” — exact numbers from PDF; mechanism more precisely described (cytochrome c retention confirmed in Fig 6C/D)
  2. Karbowski 2002 MFN2 co-localization added: body text only mentioned BAX/DRP1 co-localization; paper’s title and main finding is that MFN2 also co-localizes with BAX foci — 92% ± 10.5% of Mfn2 clusters (n=847) co-localized with Bax, 68% ± 10% of Bax clusters (n=1,238) with Mfn2. Added to Apoptosis crosstalk Stage 1.
• DOI correction confirmed: Karbowski 2002 DOI 10.1083/jcb.200209124 is correct; seeder’s note about wrong DOI 10.1083/jcb.200204140 verified — page already carried the corrected DOI; footnote now annotated with confirmation.
• footnotes updated: frank2001 and karbowski2002 moved from “pending” to “verified” with PDF details.
• ⚠️ banner removed; verified: true; scope documented in verified-scope field.
• downstream propagation needed (main agent):
  • mitofusins.md (drafted R10c) — Karbowski 2002 finding that MFN2 co-localizes with BAX foci at apoptotic fission sites is directly relevant to the MFN2 apoptosis-crosstalk section; consider adding Karbowski 2002 citation there
  • No other atomic pages appear to duplicate the specific numerical claims corrected here

[2026-05-05] verify — molecules/proteins/b2m.md

Pages verified: 1

• molecules/proteins/b2m.md — 6 substantive corrections; verified: true (partial scope — canonical-DB identity fields NCBI Gene/HGNC/Ensembl not independently re-queried)

Sources checked (all verified against local PDFs):

• doi:10.1038/nm.3898 (Smith 2015, Nat Med 21(8):932–937) — downloaded from PMC4529371; read end-to-end (8 pages incl. methods)
• doi:10.1038/329506a0 (Bjorkman 1987, Nature 329:506–512) — local PDF; read end-to-end (4 pages); DOI 329506a0 confirmed correct
• doi:10.18632/aging.101684 (Lu 2019, Aging) — local PDF; key sections verified (Fig. 1 flowchart, Table 1, Stage 1/2 architecture)
• doi:10.1186/s13041-020-0559-8 (Zhong 2020, Mol Brain 13:20) — downloaded from PMC7023753; read end-to-end (6 pages + figures)
• doi:10.1038/nature10357 (Villeda 2011, Nature 477:90–94) — downloaded from PMC3170097; read end-to-end (5 pages)
• UniProt P61769 — REST API query for disulfide bond coordinates (Cys25–Cys80 confirmed)

Corrections applied:

1. Disulfide bond WRONG — “Cys45–Cys100” → “Cys25–Cys80 in precursor numbering (positions 5 and 60 of the mature 99 aa chain)”; confirmed by UniProt P61769 REST API and Bjorkman 1987 Fig. 3 secondary structure diagram
2. B2M dose WRONG by 1000-fold — “100 mg/kg” → “100 μg/kg” (micrograms not milligrams); Smith 2015 Methods section explicit: “100 ug/kg” via intraorbital injection
3. Anti-B2M antibody arm FABRICATED — Smith 2015 does not include a systemic anti-B2M antibody treatment arm; this finding appears nowhere in the paper; removed the fabricated bullet; replaced with accurate summary of what the paper shows (genetic B2m deletion in 17-month aged mice)
4. MHC-I independence claim INVERTED — wiki stated effects are “independent of MHC-I antigen presentation per se”; Smith 2015 shows the opposite: Tap1-/- mice (lacking surface MHC-I) are PROTECTED from B2M-induced cognitive/neurogenic impairment, implicating MHC-I as a required mediator; corrected throughout
5. Mechanism section expanded with primary data — Added Zhong 2020 TLR4/NF-κB axis with verified quantitative data (n=10/subgroup behavioral; n=6/subgroup molecular; IL-1β/TNF-α elevations; BrdU+/NeuN+/DCX+ neurogenesis counts); clarified that MHC-I (Smith 2015) and TLR4 (Zhong 2020) axes both have experimental support but their relationship is unresolved
6. Footnotes substantially upgraded — Smith 2015 footnote: added exact dose (100 μg/kg), aged mouse age (17 months for KO arm, 18 months for other arms), sex (all male), key quantitative results, and Tap1-/- finding; Zhong 2020: added n values, behavioral test (MWM not RAWM), cytokine and neurogenesis quantitation; Villeda 2011: added full factor list (CCL2, CCL11, CCL12, CCL19, Haptoglobin, B2M), volume/journal info; Bjorkman 1987: DOI correction confirmed, full author list added

GrimAge component confirmed: B2M is one of 7 plasma protein DNAm surrogates (Lu 2019 Fig. 1 explicit); list matches grimage-2019.md verified page; no corrections needed to GrimAge section.

Villeda 2011 context confirmed: B2M listed by name in paper text as one of 6 factors elevated across both normal aging and heterochronic parabiosis conditions; CCL11 was primary candidate functionally tested.

Downstream pages for main agent propagation:

• No pages currently carry Smith 2015 as a footnote source for specific factual claims; corrections are self-contained within b2m.md
• If any intervention page cites “Smith 2015 anti-B2M antibody treatment in aged mice” — that claim is fabricated and must be removed

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

Pages verified: 1

• molecules/proteins/klotho.md — 6 corrections (see below); verified: true (partial scope — Kurosu 2005, Kuro-o 1997, Arking 2005, Liu 2007 unverifiable)

Sources checked:

• doi:10.1073/pnas.022484299 (Arking 2002 PNAS) — downloaded during session (bronze OA via camoufox); PDF verified end-to-end
• doi:10.1016/j.celrep.2014.03.076 (Dubal 2014 Cell Reports) — downloaded during session (gold OA via camoufox); PDF verified end-to-end
• doi:10.3389/fcvm.2020.617842 (Saar-Kovrov 2021 Front Cardiovasc Med) — downloaded during session (gold OA via PMC S3); PDF verified end-to-end
• doi:10.1126/science.1112766 (Kurosu 2005 Science) — green OA (PMC2536606) but PDF not downloadable (archive failed; PMC PDF endpoint returns HTML redirect; retry needed)
• doi:10.1038/36285 (Kuro-o 1997 Nature) — permanently closed-access; unverifiable
• doi:10.1161/01.RES.0000157171.04054.30 (Arking 2005 Circ Res) — permanently closed-access; unverifiable
• doi:10.1126/science.1143578 (Liu 2007 Science) — permanently closed-access; unverifiable

Supplementary checks:

• GenAge HAGRID 17 for KL — confirmed via HAGR web (Build 21)
• Open Targets ENSG00000133116 druggability — confirmed tier 3 via GraphQL API

Corrections:

1. Arking 2002 cohort: “525 Ashkenazi Jewish individuals” → three-population study: Bohemian Czechs (n=805), Baltimore Caucasians (n=1,143), Baltimore African-Americans (n=468) — Ashkenazi Jewish cohort does not appear anywhere in the paper
2. Arking 2002 effect size: “~3-year increase in life expectancy” → removed; paper reports no such absolute figure; corrected to 1.43-fold (95% CI 1.02–2.01) survival advantage to age 75 (Czech cohort, P=0.036)
3. Dubal 2014 age range: “53–86” → “52–85” (paper’s stated inclusion range)
4. Dubal 2014 description: expanded with verified numerics: n per cohort breakdown, Cohen’s d=0.34, MMSE≥28 inclusion criterion, mouse experiment details (probe trial 2.5x faster, p<0.01 log rank for survival)
5. Population caveat: removed claim that “both Arking studies used Ashkenazi Jewish cohorts exclusively” — Arking 2002 used no Ashkenazi cohort; Arking 2005 population not independently verifiable (closed-access)
6. Druggability-tier: null → 3 (Open Targets API confirmed; no approved drug, biologic surface accessibility predicted)

Liu 2007 attribution (10.1126/science.1143578 = “Augmented Wnt Signaling in a Mammalian Model of Accelerated Aging”, Liu H. et al.) — confirmed correct via Crossref.

Pages unverifiable (closed-access):

• Kuro-o 1997 — not_oa; discovery/phenotype claims remain unverified against full text
• Arking 2005 — not_oa; cardiovascular claims remain unverified against full text
• Liu 2007 — not_oa; Wnt-antagonism mechanism claims remain unverified

Downstream pages to check (main agent):

• studies/arking-2002-klotho-human-aging.md — if this study page exists, its cohort description and effect size are likely wrong; propagate corrections
• studies/kurosu-2005-klotho-suppression-aging.md — if this study page exists, flag that lifespan percentages still need PDF verification
• Any MOC or hallmark page citing the “525 Ashkenazi Jewish” or “+3 year life expectancy” claim from Arking 2002 should be corrected

[2026-05-05] verify | molecules/proteins/ccl11.md

Pages verified: 1

• molecules/proteins/ccl11.md — 9 corrections; verified: true (partial scope — Villeda 2014 PDF undownloadable; canonical-DB identity fields not re-verified)

Sources checked:

• doi:10.1038/nature10357 (Villeda 2011, Nature) — 17-page PMC PDF read end-to-end; verified
• doi:10.1084/jem.179.3.881 (Jose 1994, J Exp Med) — 7-page PDF read end-to-end; verified
• doi:10.1038/nm.3898 (Smith 2015, Nat Med) — 5-page PMC PDF read end-to-end; verified
• doi:10.3389/fnagi.2017.00402 (Hoefer 2017, Front Aging Neurosci) — 8-page PDF read end-to-end; verified
• doi:10.1007/s11357-020-00297-8 (Mehdipour 2021, GeroScience) — 6 of 18 pages read; verified for CCL11-relevant claims
• doi:10.3389/fimmu.2015.00084 (Williams 2015, Front Immunol) — 3-page PDF read end-to-end; is a historical opinion piece, not a structural biochemistry source
• doi:10.1038/nm.3569 (Villeda 2014, Nat Med) — green OA but PDF download failed (no retrievable URL survives archive filtering); no-fulltext-access

Corrections:

1. Proteomics screen numbers: “~50 proteins elevated in old blood” → 66 measured; 17 correlated with neurogenesis decline; 15 elevated in heterochronic parabionts; 6-factor intersection priority list (CCL2, CCL11, CCL12, CCL19, Haptoglobin, B2M)
2. CCL11 fold-change “~6-fold” flagged — not stated numerically in text; visual estimate from Fig 3b; tagged needs-replication
3. Human cohort correlation data added: plasma r=0.40, p=5.6e-7, 95% CI 0.26-0.53 (20-90 yr); CSF p<0.01 (20-45 vs 65-90 yr)
4. Administration route corrected: “subcutaneous pump” → “intraperitoneal injection days 1, 4, 7, 10 (10 ug/kg)”; n per group added (n=6-10 neurogenesis; n=12-16 behavioral)
5. Plasma injection n added: n=8 young-plasma, n=7 old-plasma recipients
6. Williams 2015 structural citation flagged as unsuitable — opinion piece lacks disulfide bond structural data; tagged unsourced on the structural claims
7. Hoefer 2017 quantified: FFP n=168 (median 69.4 pg/ml), EC n=160 (median 42.7 pg/ml), PC n=8 (3/8 detectable); Pearson r for age: FFP r=0.349 p=3.8e-5; EC r=0.438 p=6.8e-9
8. Mehdipour 2021 corrected: mouse ages (22-24 months), n per group (n=4-7), NBE vs ABT-263 clarified (ABT-263 did not enhance neurogenesis; weaker neuroinflam effect), CCL11 mechanism attribution qualified
9. druggability-tier: null -> 1 (Open Targets ENSG00000172156: Approved Drug labels at SM, AB, PR, OC modalities)

Pages unverifiable: Villeda 2014 — green OA but download failed; no-fulltext-access noted on page and footnote

Downstream propagation needed (main agent):

• Any study page for Villeda 2011 (if seeded) — carry corrected proteomics numbers (17/15/6, not “~50”)
• molecules/proteins/b2m.md — one of the 6-factor intersection; may carry old screen numbers; check
• hallmarks/chronic-inflammation.md and hallmarks/altered-intercellular-communication.md — may carry CCL11 claims; check against corrected Villeda 2011 numbers

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

Pages verified: 1

• molecules/proteins/terc.md — 7 corrections; verified: true (partial scope — Blasco 1997 body claims unverified; Feng 1995 and Mitchell 1999 permanently closed-access; canonical-DB identity fields not independently re-verified)

Sources checked:

• doi:10.1038/nature09603 (Jaskelioff 2011 Nature) — local PDF verified end-to-end
• doi:10.1038/35096585 (Vulliamy 2001 Nature) — local PDF verified end-to-end
• doi:10.1056/NEJMoa066157 (Armanios 2007 NEJM) — downloaded during session (bronze OA via camoufox); PDF verified end-to-end
• doi:10.1016/s0092-8674(01)80006-4 (Blasco 1997 Cell) — DOI confirmed correct via Crossref (title: Telomere Shortening and Tumor Formation by Mouse Cells Lacking Telomerase RNA; Cell 91(1):25-34, 1997); body claims not verified against PDF
• doi:10.1126/science.7544491 (Feng 1995 Science) — permanently not_oa; no-fulltext-access
• doi:10.1038/990141 (Mitchell 1999 Nature) — permanently not_oa; no-fulltext-access

Identity checks:

• GenAge human ID 7 confirmed for TERC via HAGR CSV download (row: 7,TERC,telomerase RNA component,7012)
• TERC mature length 451 nt confirmed via NCBI RefSeq NR_001566.3
• RNAcentral URS000075C808 not confirmed as hTR (API returned length 2364); corrected in Limitations

Corrections applied:

1. Jaskelioff 2011 model: “Tert-ERTA2/ERTA2 x Terc-/- background” corrected to TERT-ER knock-in at endogenous TERT locus; G4 TERT-ER homozygous mice intercrossed 4 generations
2. Jaskelioff 2011 drug: “tamoxifen” corrected to “4-hydroxytamoxifen (4-OHT)” — 2.5 mg slow-release subcutaneous pellets; tamoxifen not mentioned in paper
3. Jaskelioff 2011 quantitative results added: median survival G4 vehicle 43.5 vs G0 86.8 weeks (P<0.0001); brain weight 77.3+/-3.3% recovered to 89.7+/-4.0% of G0 (p=0.0004); survival improvement p<0.005
4. Jaskelioff 2011 cancer caveat: “cancer was not assessed long-term” was wrong; carcinogenesis was assessed, not observed over 4-week course (data not shown per paper)
5. Armanios 2007 prevalence: “~8-15%” corrected to “8% (6/73 probands)”; 5 hTERT + 1 hTR breakdown added; telomere length p-values added
6. Vulliamy 2001 family count: one family corrected to three (DCR101/DCR063/DCR082) with distinct mutations; TRAP assay finding (no significant activity difference in lymphocyte lines) added
7. RNAcentral URS000075C808 corrected in Limitations — entry returned length 2364, inconsistent with 451 nt hTR; NR_001566 recommended as stable identifier

Unverifiable claims remaining:

• Feng 1995, Mitchell 1999: permanently closed-access; 451 nt confirmed independently via NCBI RefSeq
• Blasco 1997: DOI confirmed; body claims pending PDF verification

Downstream pages to check (main agent):

• studies/jaskelioff-2011-telomerase-reactivation.md — model and drug correction needed
• studies/vulliamy-2001-terc-dyskeratosis.md — three families not one
• studies/armanios-2007-terc-ipf.md — 8% not ~8-15%
• hallmarks/telomere-attrition.md — check for stale prevalence and model-description claims

[2026-05-05] verify — molecules/proteins/dna-pkcs.md

Pages verified: 1

• molecules/proteins/dna-pkcs.md — 4 corrections; verified: true (partial scope — Bosma 1983 and Davis 2014 permanently closed-access; canonical-DB identity fields not independently re-verified)

Sources checked:

• doi:10.1016/j.cmet.2017.04.008 (Park 2017 Cell Metabolism) — downloaded during session (bronze OA via camoufox); PDF verified end-to-end; 16 pages including STAR methods
• doi:10.1016/bs.mie.2017.04.001 (Chirgadze/Sibanda 2017 Methods in Enzymology) — downloaded during session (green OA via PMC5495172); PDF verified end-to-end
• doi:10.1038/301527a0 (Bosma 1983 Nature) — permanently not_oa; SCID phenotype claims not verifiable from full text; DOI confirmed correct via DOI lookup
• doi:10.1016/j.dnarep.2014.02.020 (Davis 2014 DNA Repair) — permanently not_oa; NHEJ mechanism claims not verifiable from full text; DOI confirmed correct via DOI lookup

Corrections applied:

1. Inhibitor name: “(compound not commercially named in the Park 2017 paper)” → “NU7441 (40 mg/kg/day oral gavage, up to 12 weeks)” — NU7441 is named extensively throughout the paper including STAR methods; this was a fabricated claim of anonymity
2. Genetic model: “Genetic reduction of DNA-PKcs activity (hypomorphic allele)” → two explicit models: SCID mice (B6.CB17-Prkdc^scid/SzJ leaky nonsense mutation) and muscle-specific conditional KO (MDPKO, DNA-PKcs^fl/fl x MCK-Cre); the term “hypomorphic allele” does not appear in the paper
3. n values: footnote “~8-12/group” → per-experiment ranges: n=10/group (HFD weight/glucose); n=8/group (treadmill); n=3-6/group (Western blot quantification); n=4-5/group (rhesus CR); lower-n experiments were systematically missed by the original approximation
4. Structural method: “A 2017 cryo-EM structure” → “A 2017 X-ray crystallography study (selenomethionine-labeled; 4.3 A resolution)” — Chirgadze/Sibanda 2017 is an X-ray crystallography paper (ESRF beamline ID29/ID23); cryo-EM is not used in the paper

Unverifiable claims remaining:

• Bosma 1983 SCID phenotype details — permanently not_oa; no-fulltext-access in footnote
• Davis 2014 NHEJ mechanism details — permanently not_oa; no-fulltext-access in footnote
• Romesser 2024 and Samuels 2024 peposertib trial claims — not verified during this pass

Downstream propagation needed (main agent):

• studies/park-2017-dna-pk-metabolism-aging.md — update inhibitor name to NU7441; genetic model to SCID + MDPKO (not “hypomorphic”); n values per experiment
• studies/sibanda-2017-dna-pkcs-structure.md — update method from cryo-EM to X-ray crystallography (4.3 A)
• Any entity page citing the DNA-PKcs -> HSP90alpha -> LKB1 -> AMPK axis from Park 2017 — confirm inhibitor name is NU7441 and genetic models are SCID/MDPKO

[2026-05-05] verify — pathways/non-homologous-end-joining.md

Pages verified: 1

• pathways/non-homologous-end-joining.md — 4 corrections; verified: true (partial scope — Mahaney 2009 PDF unavailable; Steps 2-3 autophosphorylation and ligation-complex details unverified)

Sources checked:

• doi:10.1038/35088000 (Walker 2001, Nature) — 8-page PDF verified end-to-end
• doi:10.1016/s0960-9822(99)80311-x (Riballo 1999, Current Biology) — 4-page PDF + 2-page supplementary verified end-to-end
• doi:10.1042/BJ20080413 (Mahaney 2009, Biochemical Journal) — download failed (green OA, no viable URL); no-fulltext-access tagged on page and footnote

Corrections:

1. Ku ring nucleolytic-protection claim over-attributed to Walker 2001: Walker is a structural paper and does not biochemically demonstrate nuclease protection; removed unsupported attribution; tagged unsourced with note that Feldmann 2000 (Nucleic Acids Res 28:2585) is the relevant biochemical source
2. LIG4 syndrome disease table: “Microcephaly, growth retardation, pancytopenia, radiosensitivity, lymphoma predisposition [^riballo1999]” was wrong for Riballo 1999; Riballo 1999 describes one hypomorphic patient with radiosensitivity and leukemia predisposition but NO overt immunodeficiency and near-normal V(D)J recombination; corrected entry; full syndrome features re-attributed to later case series with unsourced pending O’Driscoll 2001 citation
3. Riballo 1999 footnote: updated to reflect n=1 patient (180BR cell line, age 14 ALL, R278H LIG4 hypomorph), specify residual adenylation at high ATP, note absence of immunodeficiency
4. Walker 2001 footnote: updated with accurate structural specifics (2.5 A DNA-bound, 2.7 A apo, truncated Ku80 lacking C-terminal 19K, confirmed key structural facts, corrected local PDF path)

Unverifiable claims remaining:

• Mahaney 2009 (Steps 2-3): DNA-PKcs autophosphorylation cluster residues (ABCDE, PQR) and XRCC4-LIG4-XLF filament scaffold details

Downstream pages to check (main agent):

• molecules/proteins/ku70-ku80.md (R19) — cites Walker 2001 for structural claims; verify ring-description and Ku80-truncation detail consistency
• molecules/proteins/dna-pkcs.md (R19) — cites Mahaney-dependent autophosphorylation cluster details; flag as unverified pending Mahaney 2009 PDF access
• Any page citing Riballo 1999 for full LIG4 syndrome phenotype should be checked for the same over-attribution corrected here

[2026-05-05] verify — pathways/mismatch-repair.md

Pages verified: 1

• pathways/mismatch-repair.md — 4 corrections; verified: true (partial scope — Lynch 2009 and Le 2017 not PDF-verified; see verified-scope)

Sources checked:

• doi:10.1038/nrm1907 (Jiricny 2006, Nat Rev Mol Cell Biol 7:335–346) — local PDF verified end-to-end
• doi:10.1056/NEJMoa1500596 (Le 2015, NEJM) — downloaded during session (bronze OA via camoufox); PDF verified end-to-end
• doi:10.1056/NEJMoa2017699 (André 2020 KEYNOTE-177, NEJM) — downloaded during session (bronze OA via camoufox); PDF verified end-to-end
• doi:10.1126/science.aan6733 (Le 2017, Science) — permanently not_oa; no-fulltext-access retained on footnote
• doi:10.1111/j.1399-0004.2009.01230.x (Lynch 2009, Clin Genet) — download status: pending; not verified

Corrections applied:

1. MutLβ subunit composition: “MLH1 + MLH3” → “MLH1 + PMS1” — Jiricny 2006 Table 1 explicitly lists MutLβ = MLH1 + PMS1 (function unknown); MLH1 + MLH3 is MutLγ, not MutLβ
2. MutLβ/MutLγ function description: separated into two complexes per Jiricny Table 1; MutLγ (MLH1+MLH3) correctly attributed as meiotic + MutLα backup; MutLβ (MLH1+PMS1) function noted as “currently unknown”
3. MutSβ substrate range: “Larger IDLs (2–13 nt); D-loops” → “Larger IDLs (>2 nt)” with needs-replication — Jiricny 2006 does not state the 2–13 nt upper bound or list D-loops as a MutSβ substrate per Table 1
4. key-nodes frontmatter: added “pms1” — PMS1 is the MutLβ partner per Jiricny; was absent

Footnote updates:

• Le 2015: download status updated to local path; ORR and PFS figures added; study-design corrected from “phase 2 RCT” to “phase 2” (non-randomized three-cohort design)
• André 2020: download status updated to local path; P=0.0002 added; per-arm n clarified (153 pembro, 154 chemo)
• Auto-extraction banner removed

Unverifiable claims remaining:

• Lynch syndrome gene proportion table and lifetime CRC/endometrial cancer risk figures — Lynch 2009 PDF not downloaded; figures not PDF-verified
• Le 2017 ORR 53% across 12 tumor types — permanently closed-access (#gap/no-fulltext-access)
• Aging-specific MMR attrition claims — unsourced; no primary source available
• “MSI-low aging colon” (Iyer 2008, DOI 10.1093/gerona/63.7.692) — DOI returns 404; not in archive; do not cite

Downstream pages to check (main agent):

• Any page citing mismatch-repair for MutLβ subunit composition — now correctly MLH1+PMS1
• mlh3 stub — links to MutLγ role; confirm not MutLβ
• pms1 stub — add note that PMS1 is MutLβ partner with unknown somatic MMR function

[2026-05-05] verify — molecules/proteins/brca1.md

Pages verified: 1

• molecules/proteins/brca1.md — 4 substantive corrections; verified: true (partial scope — Foulkes 2014 NEJM DOI confirmed non-existent; other primary sources not PDF-verified on this pass; canonical DB identifiers not re-checked)

Sources checked:

• doi:10.1038/nature03443 (Bryant 2005 Nature) — local PDF verified end-to-end
• doi:10.1038/nature03445 (Farmer 2005 Nature) — local PDF verified end-to-end
• doi:10.1056/NEJMra1306063 (stated as Foulkes 2014 NEJM) — confirmed non-existent: HTTP 404 at doi.org; absent from Crossref, PubMed DOI-field lookup, Europe PMC, Semantic Scholar, and all 30 Foulkes WD NEJM publications indexed in PubMed

Corrections:

1. Bryant 2005 footnote cell lines: “Capan-1, V-C8, MEFs” → “V79, V-C8 (BRCA2-deficient), MCF7, MDA-MB-231, SW480SN.3” — Capan-1 and MEFs are not Bryant 2005 cell lines
2. Farmer 2005 footnote cell lines: “MDA-MB-436, EUFA423” → “mouse ES cells with homozygous Brca1 or Brca2 deletion (isogenic WT controls)” — those human lines do not appear in Farmer 2005
3. Farmer 2005 footnote compounds: “olaparib precursor compound” removed; named as KU0058684 (SF50=35 nM BRCA1-KO vs ~2 uM WT; 57-fold selectivity) and KU0058948; in-vivo p-values added
4. PARP mechanism body text: Bryant 2005 specificity for BRCA2-deficient cells clarified; Farmer 2005 BRCA1-null ES data noted as “unpublished” at time of publication
5. Foulkes DOI: all claims tagged no-fulltext-access; lifetime risk table flagged unreliable; candidate replacement citations offered (Antoniou 2003 PMID 12677558; Chen & Parmigiani 2007 PMID 17452630)

Unverifiable claims:

• Lifetime breast/ovarian/contralateral cancer risk percentages — invalid source DOI; cannot be confirmed
• ESR1/triple-negative mechanistic link — also sourced to invalid DOI
• Other primary sources on page not PDF-verified on this pass

Downstream pages to check (main agent):

• studies/bryant-2005-parp-brca2-synthetic-lethality.md — cell lines correction needed
• studies/farmer-2005-parp-brca-targeting.md — cell lines + compound name correction needed
• pathways/homologous-recombination.md — confirm 2026-05-05 HR-page verification pass already fixed Bryant/Farmer footnotes
• Any page citing BRCA1 lifetime cancer risk percentages — figures currently unsourced

[2026-05-05] verify — molecules/proteins/ercc1.md

Pages verified: 1

• molecules/proteins/ercc1.md — 10 corrections; verified: true (partial scope — Gregg 2011 review and Deng 2017 closed-access, not_oa, unverified; canonical-DB identity fields not re-checked)

Sources checked:

• doi:10.1038/nature05456 (Niedernhofer 2006 Nature) — PDF verified end-to-end; locally available
• doi:10.1038/nature19329 (Vermeij 2016 Nature) — PDF downloaded 2026-05-05 and verified end-to-end
• doi:10.1016/s0960-9822(06)00190-4 (Weeda 1997 Current Biology) — PDF downloaded 2026-05-05 and verified end-to-end
• doi:10.1086/512486 (Jaspers 2007 Am J Hum Genet) — PDF downloaded 2026-05-05 and verified end-to-end
• doi:10.1038/ng1193-217 (McWhir 1993 Nature Genetics) — DOI confirmed correct via Crossref; not OA; PDF not in archive
• doi:10.1016/j.dnarep.2011.04.026 (Gregg 2011 DNA Repair) — closed-access (not_oa); unverified
• doi:10.1016/j.jflm.2017.05.005 (Deng 2017 J Forensic Leg Med) — closed-access (not_oa); unverified

Corrections:

1. Mouse model identity (Niedernhofer 2006): “Ercc1-/delta allelic series — compound heterozygote” → “Ercc1-/- knockout”; the compound heterozygote Ercc1^Δ/− is the Vermeij 2016 model; Niedernhofer 2006 uses Ercc1-/- nulls
2. “retains ~10% of normal ERCC1 protein level” removed from Niedernhofer section — this ~15% transcript figure is from Weeda 1997 (Ercc1*292 allele), not Niedernhofer 2006
3. Niedernhofer 2006 progeroid feature list corrected: “loss of cerebellar Purkinje cells” removed (not in Niedernhofer 2006; this is Vermeij 2016); “elevated glucocorticoids” removed (not in paper); “hematopoietic failure (anemia, reduced HSC self-renewal)” simplified to “bone marrow abnormalities” per paper; added paper-attested features: dystonia/ataxia, kyphosis, renal insufficiency; somatotropic axis data given with exact p-values (IGF-1 P<0.001, insulin P<0.03, glucose P<0.001)
4. Niedernhofer 2006 mechanistic framing: “endogenous oxidative DNA damage” → “endogenous DNA damage (particularly cytotoxic interstrand crosslinks)” per paper’s own hypothesis
5. Vermeij 2016 section fully rewritten with verified numbers: males 10→35 wk (250%; p<0.0001), females 13→39 wk (200%; p<0.0001); replication cohort 180% (p<0.0001); ~50% neuron retention; γH2AX foci reduction in Purkinje cells; Xpg-/- extension data added; DR protocol corrected (week 7 start at 10%, reaching 30% by week 9 — not “from weaning”)
6. Vermeij 2016 hedge “exact figures require PDF verification” removed; “(single study)” gap tag updated to specify intra-group replication only
7. Jaspers 2007 NER residual activity: “20–30% residual activity” → “UDS (GG-NER) 15% of normal; TC-NER 13% of normal”
8. Jaspers 2007 patient description: “cataracts” → “bilateral microphthalmia and blepharophimosis” (cataracts not reported); mutations specified (Q158X + F231L); age at death corrected to 14 months; cause of death (respiratory failure) added
9. McWhir 1993 DOI confirmed correct (10.1038/ng1193-217); “ng1093-217” alternate form resolves to nothing
10. All three newly-downloaded footnotes updated from “download status: pending” to “PDF locally available; verified”; Weeda 1997 model description corrected to Ercc1*292 truncation specifics

Unverifiable claims (partial scope):

• Gregg 2011-derived structural claims (HhH2 domain residues, incision positions, NER mechanism details): closed-access; tagged with [^gregg2011] but not verified against PDF
• Deng 2017 PBMC correlations (r=−0.578, r=−0.844): closed-access; numbers match original extraction but not verified against PDF

Downstream pages needing update:

• pathways/nucleotide-excision-repair.md — already updated in 2026-05-05 NER pass; cross-check that Ercc1-/- vs Ercc1^Δ/− distinction is consistent with corrections here
• Any page citing Vermeij 2016 lifespan figures — verify ~200% hedge is resolved to sex-specific data (males 250%, females 200%)

[2026-05-05] verify — molecules/proteins/ogg1.md

Pages verified: 1

• molecules/proteins/ogg1.md — 10 corrections; verified: true (partial scope — Visnes 2018 and Sheng 2012 are not_oa; canonical-database identity fields not re-checked against databases)

Sources checked:

• doi:10.1073/pnas.96.23.13300 (Klungland 1999 PNAS) — PDF downloaded and verified end-to-end
• PMID 11454679 (de Souza-Pinto 2001 Cancer Research) — verified via PubMed abstract (no DOI registered; pre-DOI era)
• doi:10.1158/0008-5472.can-06-4802 (Sakamoto 2007 Cancer Research) — PDF downloaded and verified end-to-end
• doi:10.1038/s41467-020-16361-y (Pao 2020 Nature Communications) — PDF downloaded and verified end-to-end
• doi:10.1002/cmdc.202200310 (Wallner 2022 ChemMedChem) — PDF downloaded and verified end-to-end
• doi:10.1126/science.aar8048 (Visnes 2018 Science) — not_oa; no local PDF; TH5487 claims unverified against PDF
• doi:10.1172/JCI65053 (Sheng 2012 JCI) — not_oa; no local PDF; neurodegeneration claims unverified against PDF

Corrections:

1. “Sakumi et al. 2007” → “Sakamoto et al. 2007” throughout (first author is Katsumi Sakamoto; Kunihiko Sakumi is a middle author — confirmed from PDF title page)
2. Sakamoto 2007 body: paper studies MUTYH single-KO only; DKO cancer phenotype (lung, ovarian, lymphomas) correctly re-attributed to Xie et al. 2004 (Cancer Res 64:3096)
3. Sakamoto 2007 body: n added (121 Mutyh-/- + 109 WT); intestinal tumor data (10/121 vs 0/109; P < 0.001); age (18 months); background (near-congenic C57BL/6J)
4. Sakamoto 2007 footnote: first author corrected; n and key stats added; “download pending” → ""
5. Klungland 1999 body: ES cell origin specified (129SV → C57BL/6J blastocysts); 8-oxoG fold-change specified as ~1.7x in liver nuclear DNA (HPLC-ECD; n=5/genotype); mutation frequency specifics added (2-3x liver, not testis); histopathology observation window (oldest 18 months, two sacrificed at 8 and 11 months) added
6. Klungland 1999 footnote: volume/pages added; n stated; “download pending” → ""
7. Pao 2020 body: AD model specified as 5XFAD; mechanism made precise (HDAC1 deacetylates OGG1 reversing p300 acetylation, stimulating AP-lyase); exifone dose (50 mg/kg/day IP, 4 weeks), target age (17 months) added
8. Pao 2020 footnote: model, mechanism, dose, and outcome details added; “download pending” removed
9. Wallner 2022 body: TH8535 IC50 200 nM stated; selectivity panel result and co-crystal structure at 2.45 Å added
10. Wallner 2022 footnote: IC50, selectivity, cancer viability, and structural data added
11. de Souza-Pinto 2001 footnote: mouse background (C57BL) added; journal volume/pages added; verification method stated
12. Gaps table: stale “Sakumi 2003 → DKO” entry replaced with accurate source-tracking entries for Lu 2008 DOI correction and Xie 2004 DKO gap
13. ⚠️ banner removed; verified flag flipped to true

Downstream pages to check (main agent):

• molecules/proteins/mutyh — correct “Sakumi et al. 2007” → “Sakamoto et al. 2007” if cited; confirm DKO phenotype attributed to Xie 2004
• Any study page for sakumi-2007 if it exists — update first-author attribution

[2026-05-05] verify — pathways/base-excision-repair.md

Pages verified: 1

• pathways/base-excision-repair.md — 9 corrections; verified: true (partial scope — Lindahl 1993 closed-access; Polβ 39 kDa weight claim unverified against that source)

Sources checked:

• doi:10.1101/cshperspect.a012583 (Krokan & Bjoras 2013 Cold Spring Harb Perspect Biol) — full PDF verified end-to-end (18 pages)
• doi:10.1073/pnas.96.23.13300 (Klungland et al. 1999 PNAS) — full PDF verified end-to-end
• doi:10.1080/15548627.2016.1257467 (Fang & Bohr 2016 Autophagy) — full PDF verified (2-page autophagic punctum)
• doi:10.1016/j.dnarep.2008.03.011 (de Souza-Pinto et al. 2008 DNA Repair) — verified via PMC full text (PMC4625841)
• doi:10.1016/j.neurobiolaging.2008.07.004 (Gredilla et al. 2010 Neurobiol Aging) — verified via PMC full text (PMC2858237)
• doi:10.1016/j.neurobiolaging.2010.06.019 (Gredilla et al. 2012 Neurobiol Aging) — verified via PMC full text (PMC3041866)
• doi:10.1016/j.mad.2012.01.005 (Hegde et al. 2012 Mech Ageing Dev) — verified via PMC full text (PMC3351531)
• doi:10.1016/j.mad.2009.05.003 (Soerensen et al. 2009 Mech Ageing Dev) — verified via PMC full text (PMC2764052)
• doi:10.1038/362709a0 (Lindahl 1993 Nature) — closed-access (not_oa); no-fulltext-access applied

Corrections:

1. Long-patch BER resynthesis length “2–12 nt” → “2–10 nt” (Krokan 2013 p.8 explicitly: “a gap of 2–10 nucleotides”)
2. MUTYH glycosylase classification: removed from bifunctional category — Krokan 2013 Table 1 marks MUTYH as monofunctional (“M”); it excises adenine mispaired with 8-oxoG but has NO lyase activity; note added to Step 1 section
3. PARP1 role in BER overstated: removed claim that PARP1 “recruits the XRCC1–LIG3 ligation complex to the nick” — Krokan 2013 explicitly states PARP1 role in short-patch BER is “unsettled”; PARP1 is properly characterized as a SSBR sensor with overlap at nick intermediates, not a canonical BER co-factor for all lesion types
4. LIG1 vs LIG3 in short-patch BER: updated table and Step 4 text — LIG1 is now the major nuclear ligase for short-patch BER (per Krokan 2013 citing Simsek 2011, Gao 2011); LIG3 is essential in mitochondria; prior table showed LIG3 as the short-patch nuclear ligase
5. Mitochondrial BER exclusivity overstated: intro and body text changed from “the only repair pathway operating on mtDNA” → “dominant and best-characterized”; de Souza-Pinto 2008 explicitly states mitochondria also have recombination and direct damage reversal
6. XRCC1 in mitochondria: table entry changed from “Disputed; possibly absent” → “Absent from mitochondria” — de Souza-Pinto 2008 explicitly confirms XRCC1 is absent; LIG3α functions without it in the mitochondrial matrix
7. Gredilla 2010 hippocampus claim: “declines with age in mouse brain cortex and hippocampus” → cortex decline was significant (p≤0.003); hippocampus showed only minor changes; added age groups (1, 5, 10, 15–20 months)
8. Gredilla 2012 AD model claim: removed statement that BER declined in “Alzheimer’s disease models” — 3xTgAD mice showed NO significant BER changes at either pre-symptomatic or symptomatic ages; synaptosomal decline is a feature of normal aging only
9. Fang & Bohr 2016 attribution: clarified that this paper is a 2-page ATM-specific commentary, not a general BER-PARP1 review; NAD+ axis in paper operates via ATM → PARP1 → SIRT1 → mitophagy in ATM-deficient cells, not general BER→PARP1

Unverifiable claims:

• Lindahl 1993 (Polβ 39 kDa, error-prone polymerase context) — closed-access; no-fulltext-access applied to footnote
• Klungland 1999 footnote n-count: PDF confirmed; n per group not explicitly tallied per group in methods (five mice per genotype used for HPLC-ECD; two animals for MEF experiments)

Cross-page consistency notes (main agent to check):

• molecules/proteins/ogg1.md (R19 batch, verified 2026-05-05) — uses Klungland 1999; ogg1.md has correct mixed C57BL/6×129SV background; consistent with BER page; no conflicts found
• molecules/proteins/parp1.md (R19 batch, verified 2026-05-05) — cites PARP1 in BER; parp1.md correctly characterizes PARP1 as a SSBR sensor and notes BER role is unsettled; consistent with corrected BER page; no propagation needed
• MUTYH correction (monofunctional reclassification): check if molecules/proteins/mutyh.md describes MUTYH as bifunctional; if so, that page needs correction to match Krokan 2013

[2026-05-05] verify — R20 batch (13 pathway/protein synthesis pages, lightweight sweep)

Pages flipped to verified: true (all 13):

• pathways/notch-pathway.md
• pathways/wnt-beta-catenin.md
• pathways/tgf-beta.md
• pathways/bmp-signaling.md
• pathways/hedgehog-pathway.md
• pathways/hippo-yap-taz.md
• pathways/integrated-stress-response.md
• molecules/proteins/smad2-smad3.md
• molecules/proteins/yap-taz.md
• molecules/proteins/eif2alpha.md
• molecules/proteins/atf4.md
• molecules/proteins/perk.md
• molecules/proteins/gcn2.md

Verification scope (per CLAUDE.md verified-scope: partial-verification convention): DOI sanity-check via. Of ~75 unique cited DOIs across the batch, 73 had DOI lookup titles matching the cited author/year/topic. Two DOIs unresolved in archive but already documented in the source pages: 10.1038/8484 (Tao 1999, Hippo) and 10.1073/pnas.2518812123 (Götz 2026, GCN2 fly). One DOI confirmed-mismatch and explicitly flagged in body: 10.1126/science.273.5275.622 (Marigo 1996 PTCH cloning) → resolves to a prion paper, hedgehog page leaves as needs-canonical-id. Cross-link integrity confirmed against verified atomic pages (satellite-cells, hematopoietic-stem-cells, klotho).

Carlson 2008 DOI conflict resolution:

• Confirmed via DOI lookup: 10.1038/nature07034 = “Imbalance between pSmad3 and Notch induces CDK inhibitors in old muscle stem cells” (Carlson, Hsu, Conboy 2008 Nature) — CORRECT.
• Confirmed via DOI lookup: 10.1038/nature06849 = “Free choice activates a decision circuit between frontal and parietal cortex” (Desmurget Nature — neuroscience, NOT Carlson) — INCORRECT.
• smad2-smad3.md uses correct DOI (10.1038/nature07034); banner note retained in verified-scope for posterity.

Corrections applied (cross-page propagation):

1. Added Carlson 2008 (10.1038/nature07034) citation to pathways/tgf-beta.md — previously cited only Carlson 2009 (Aging Cell, 10.1111/j.1474-9726.2009.00517.x); the foundational mechanistic paper was missing. Now cited alongside Carlson 2009 in the satellite cells aging section, with cross-link to smad2-smad3 for molecular detail.
2. Fixed broken wikilink [[tgf-beta-pathway]] → [[tgf-beta]] in pathways/hippo-yap-taz.md (line 212) and molecules/proteins/yap-taz.md (line 140). Previously tagged as stub but the actual file pathways/tgf-beta.md exists (R20 seeded).
3. Fixed broken wikilink [[wnt-signaling]] → [[wnt-beta-catenin]] in pathways/bmp-signaling.md (line 183).
4. Fixed broken wikilink [[perk-pathway]] → [[perk]] in molecules/proteins/atf4.md (frontmatter pathways: field and Pathway membership section). The protein page molecules/proteins/perk.md is the canonical PERK target; no pathways/perk-pathway.md file exists or should exist (PERK is a kinase node within the ISR/UPR pathways, not its own pathway).

Cross-page propagation gaps (NOT addressed; flagged for next pass):

• hallmarks/stem-cell-exhaustion.md Signaling pathways table currently lists only [[notch-pathway]] and [[mtor]] etc.; should be enriched with [[wnt-beta-catenin]], [[tgf-beta]], [[bmp-signaling]], [[hedgehog-pathway]], [[hippo-yap-taz]] now that all five are seeded and verified. Out of scope for verifier sweep; flag for R21 propagation pass.
• hallmarks/loss-of-proteostasis.md does not yet reference the ISR/PERK/GCN2/eIF2α/ATF4 axis; should be propagated. Out of scope; flag for R21.
• cell-types/satellite-cells.md and cell-types/hematopoietic-stem-cells.md may have stale claims overlapping with the new pathway pages; surface checks (grep for tgf-beta/wnt-beta-catenin/notch-pathway in these MOCs) suggest cross-references already exist but quantitative claims about niche signaling were not re-validated against the new pages. Flag for R21.