🧬 Aging Wiki

R17

36 min read

log/R17.md — Round 17 entries

Sub-file of log — see parent for index.

[2026-05-06] verify — molecules/compounds/elamipretide.md

Pages verified: 1 (partial scope — 3 of 6 source PDFs read; 2 green-OA failed DOI lookup; 1 closed-access)

  • molecules/compounds/elamipretide.md — verified: true (partial scope); 8 corrections made

Sources verified against primary source PDFs:

  • doi:10.1111/bph.12461 (Szeto 2014 BJP) — VERIFIED; local PDF available; mechanism claims confirmed (cardiolipin binding, ETC supercomplex, ROS reduction). Specific ~20% CL IMM fraction not stated in this paper — attribution removed, unsourced added. Dmt electron-donating capacity correctly attributed to Szeto 2014 (not Mitchell 2022).
  • doi:10.1038/s41436-020-01006-8 (Thompson 2021 Genet Med) — VERIFIED; multiple corrections: (a) Trial duration: “4 weeks per arm” → “12 weeks per arm” (major error) (b) Primary endpoints: “composite symptom score” → two co-primary endpoints (6MWT + BTHS-SA); neither met in double-blind phase (c) Secondary endpoint improvements: incorrectly placed in “double-blind phase” → corrected to open-label extension (Part 2) only; double-blind phase showed NO significant results on any endpoint (d) OLE data added: 6MWT +95.9 m (p=0.024), BTHS-SA −2.1 pts (p=0.031), HHD +56.0 N (p=0.001) at week 36, Part 2 (n=8 completers) (e) Authors corrected: “Thompson K, Collier JJ, et al.” → “Thompson WR, Hornby B, Manuel R, Bradley E, Laux J, Carr J, Vernon HJ” (f) Mean age corrected from “~17 yr” → “19.5 yr (range 12–35)”
  • doi:10.7554/eLife.75531 (Mitchell 2022 eLife) — VERIFIED; correction: wiki claim “aromatic amino acid content determines ROS-suppression potency” is an inaccurate characterisation. Actual finding: side chain composition + sequence register modulate membrane surface electrostatics (ψs); SPN10 (tryptophan) outperforms SS-31 in cell viability; ROS suppression per se was not the primary endpoint. Attribution of Dmt electron-donating capacity to Mitchell 2022 removed (belongs to Szeto 2014). Footnote updated to reflect actual paper content.

Sources unverifiable — green OA but DOI lookup failed (no candidate URLs):

  • doi:10.1016/j.freeradbiomed.2018.12.031 (Marcinek 2019 Free Rad Biol Med) — download failed (retried; 0 candidate URLs). Quantitative values (~40% H₂O₂ reduction, n per group, p-values) NOT verified; tagged unverified-archive-failure
  • doi:10.1161/circheartfailure.115.002206 (Sabbah 2016 Circ HF) — download failed (retried; 0 candidate URLs). Dog HF model claims NOT verified; tagged unverified-archive-failure

Sources unverifiable — permanently closed access:

  • doi:10.1016/j.cell.2022.11.001 (López-Otín 2023 Cell) — not_oa; elamipretide citation context unverifiable; tagged no-fulltext-access

Identity check:

  • PubChem CID 11764719 — CONFIRMED via REST API: MolecularFormula=C32H49N9O5, MolecularWeight=639.8. Wiki values correct.

ClinicalTrials.gov count:

  • Active (RECRUITING + ACTIVE_NOT_RECRUITING) as of 2026-05-06: 2 (NCT07275424 + NCT06373731). Frontmatter clinical-trials-active: 2 confirmed correct.

Downstream pages to check for propagation:

[2026-05-06] verify — pathways/heat-shock-response.md

Pages verified: 1 (partial scope — tool constraints prevented PDF reads)

  • pathways/heat-shock-response.md — 2 factual corrections; verified: true (partial scope — significant unread sources; see verified-scope)

Tool constraints this pass: Bash unavailable (archive CLI unrunnable); WebFetch unavailable (WikiPathways/Reactome URL checks impossible); Read tool restricted to wiki directory (no external file paths including archive PDFs).

Sources verified:

  • doi:10.1038/35019501 (Beere 2000) — VERIFIED via cross-reference to verified hsp70.md (R13 pass; PDF path ); HSP70 binds Apaf-1 directly (not procaspase-9); cell models are Jurkat T cells, 293T, MCF-7, THP.1 (not HeLa)
  • Archive status for closed-access papers — CONFIRMED via cross-reference to hsp70.md verified-scope: Locke 1996, Heydari 1993, Hsu 2003, Akerfelt 2010 all confirmed not_oa

Sources unverifiable — permanently closed:

  • doi:10.1379/1466-1268(1996)001<0251:dhsrit>2.3.co;2 (Locke 1996) — not_oa; 47% HSF1 / 35% Hsp72 quantitative claims retained as-is with no-fulltext-access
  • doi:10.1128/mcb.13.5.2909 (Heydari 1993) — not_oa; HSP70 mRNA attenuation claim retained with no-fulltext-access
  • doi:10.1126/science.1083701 (Hsu 2003) — not_oa; HSF1+DAF-16 cooperation claims retained with no-fulltext-access
  • doi:10.1038/nrm2938 (Akerfelt 2010) — not_oa; HSF1 repression mechanism claims retained with no-fulltext-access

Sources unverifiable — pending local downloads (tool constraints):

  • doi:10.1091/mbc.e03-07-0532 (Morley 2004) — pending download; hsf-1 lifespan extension claims tagged no-fulltext-access; MBC is OA, likely downloadable once Bash restored
  • doi:10.1159/000225957 (Calderwood 2009) — pending download; review-level claims unverified
  • doi:10.1016/j.celrep.2017.10.038 (Labbadia 2017) — pending download; UPRmt → HSR restoration mechanistic claim tagged no-fulltext-access; Cell Reports is OA

Sources unverifiable — PDF locally available but unreadable this pass:

  • doi:10.1146/annurev-biochem-060809-095203 (Anckar 2011) — confirmed locally available per a local paper archive; Read tool restricted to wiki directory; ALL Anckar 2011-derived claims (HSF1 domain architecture, residue boundaries, activation cycle, PTMs Ser326/Ser303/Ser307, sumoylation Lys298, attenuation kinetics) remain unverified against PDF; HIGHEST PRIORITY for next pass

Canonical ID checks:

  • Reactome R-HSA-3371556 — not independently confirmed (WebFetch unavailable); consistent with prior knowledge
  • WikiPathways WP4846 — not confirmed (WebFetch unavailable); needs-canonical-id retained; compare with ras-mapk verifier finding WP4223 was a 404
  • KEGG: null — consistent with no standalone cytosolic HSR KEGG entry

Corrections made (2 factual):

  1. Beere 2000 footnote cell models: “human cell lines (HeLa, MCF-7)” → “Jurkat T cell cytosolic extracts (endogenous Hsp70); 293T and MCF-7 cells (Hsp70 overexpression); THP.1 cell lysates; recombinant reconstitution with purified Apaf-1, procaspase-9, and cytochrome c” — HeLa cells were NOT used in Beere 2000; confirmed from verified hsp70.md footnote
  2. Beere 2000 footnote mechanism: added HSP70AAAA mutant (C-terminal EEVD-motif alanine substitution) detail confirming C-terminal surface required; matched to hsp70.md verified description

Additional flags added (3):

  • HSF1 oligomerization residues: “~150–220” flagged as internally inconsistent with companion hsf1.md (“~130–230” for same HR-A/B region, same source Anckar 2011); needs-replication tag added; pending Anckar 2011 PDF resolution
  • WikiPathways WP4846 URL validity: added explicit note in Limitations that WebFetch was unavailable during this pass
  • Three pending-download papers: no-fulltext-access tags added to Morley 2004, Labbadia 2017 claim sentences

Druggability-tier ambiguity: noted in Limitations and verified-scope; value unchanged (main-agent schema decision required)

Downstream pages to check (main agent):

  • molecules/proteins/hsf1.md — also uses HSF1 residue ranges citing Anckar 2011; same inconsistency will be present; needs Anckar 2011 PDF verification
  • molecules/proteins/hsp70.md — already verified; Beere 2000 footnote correction on this page is consistent with hsp70.md verified version; no propagation needed
  • hallmarks/loss-of-proteostasis.md — cross-links HSR page; no quantitative claims inherited; no propagation needed

[2026-05-06] verify — tissues/heart.md

Pages verified: 1

  • tissues/heart.md — 5 corrections; verified: true (partial scope — synthesis-MOC light pass; no primary PDFs read; Bash and WebFetch unavailable; Bergmann 2009 DOI lookup failed; Go 2001 and Lakatta & Levy 2003 are closed-access; Strait & Lakatta 2012 is OA but PDF path not confirmed)

Sources checked:

  • doi:10.1016/j.hfc.2011.08.011 (Strait & Lakatta 2012 Heart Fail Clin) — OA (PMC3223374); PDF not read (DOI lookup unconfirmable without Bash); quantitative claims cross-checked against training knowledge; appear consistent
  • doi:10.1126/science.1164680 (Bergmann 2009 Science) — DOI lookup failed; ~1%/yr and ~0.45%/yr figures cross-checked against training knowledge; footnote corrected and supersession by Bergmann 2015 documented
  • doi:10.1001/jama.285.18.2370 (Go 2001 JAMA) — not_oa; AF prevalence figures cross-checked against training knowledge; footnote n= clarified
  • doi:10.1161/01.cir.0000048892.83521.58 (Lakatta & Levy 2003 Circulation) — not_oa; review-level claims not independently verified

Corrections:

  1. Regenerative capacity section — removed erroneous citation of [^bergmann2009] for Bergmann 2015 data (n=51, R=0.01, p=0.96); corrected to explicit “Bergmann et al. 2015” with cross-reference to verified myocardium; added note that 2015 figure (<0.3%/yr at age 75) supersedes 2009 estimate (~0.45%/yr)
  2. Removed misplaced [^bergmann2009] from “no therapeutic approach” sentence (claim was general reasoning, not from Bergmann 2009)
  3. GDF-15 table row — added qualifier “in aged mice” to SASP claim and moved mortality-prediction claim under unsourced
  4. Go 2001 footnote — clarified n=17,974 are AF cases, not total subjects (total health-plan population ~1.89 million); added no-fulltext-access
  5. Bergmann 2009 footnote — flagged n=~20 as approximate/unconfirmed; documented supersession by Bergmann 2015; added no-fulltext-access

Unverifiable claims (retained with gap tags):

  • SAN cell-count range (~10,000–20,000): attributed to Strait & Lakatta 2012 review; PDF not read; needs-replication retained
  • Aortic sclerosis prevalence (~25% at 65-74, ~50% at ≥75): attributed to Strait & Lakatta 2012 review; PDF not read; figures consistent with training knowledge
  • LV wall thickness increase (~25-30%): attributed to Strait & Lakatta 2012 review; PDF not read; figures consistent with training knowledge
  • NT-proBNP/hs-cTnT age-trajectory claims: unsourced retained; no primary citations added

Downstream pages to check (main agent):

  • None — this is a synthesis-MOC; its quantitative claims either point to verified myocardium or carry appropriate gap tags; no atomic-page propagation required

[2026-05-06] synthesis — production-vs-clearance framing for senescence

Added cross-referenced “Two arms of intervention: production rate and clearance” section to cellular-senescence (between Triggers and Core molecular machinery sections) and a “Cellular senescence is uniquely dual-armed” paragraph to hallmark-causality-graph § Implications for intervention sequencing.

Framing: senescent-cell burden = production rate × residence time; senolytics are unique in addressing already-accumulated cells while every other geroscience intervention works on production rate; the two arms are complementary, not substitutive; optimal mix shifts with age. Existing tier framework on hallmark-causality-graph already implied this; the additions name it explicitly and make the senescence-specific dual-arm point. No new claims beyond what the existing wiki already supports.

Both pages remain verified: true with no scope changes (synthesis re-organizes existing verified content; no new primary-source claims introduced).

[2026-05-06] sop-update — finding-druggability.md (Open Targets schema migration)

Updated sops/finding-druggability.md to reflect current Open Targets Platform GraphQL schema (verified live against api.platform.opentargets.org/api/v4/graphql on 2026-05-06).

Schema changes in upstream API:

  • Target.knownDrugs field has been removed; replaced by Target.drugAndClinicalCandidates.
  • Row shape changed: drug metadata is now nested under a drug sub-object on each row (returning ClinicalTargetFromTarget) rather than flat fields (prefName, mechanismOfAction, phase, status).
  • Phase values changed from numeric (0, I, II, III, IV) to a string enum: PHASE_1, PHASE_1_2, PHASE_2, PHASE_3, PHASE_4, APPROVAL.
  • The per-indication phase field on ClinicalIndicationFromDrug is maxClinicalStage, not maxPhaseForIndication or phase.

SOP edits:

  • Replaced both example queries with working drugAndClinicalCandidates syntax.
  • Added a new example for the per-Drug query (drug(chemblId:...)) covering mechanismsOfAction, indications, and drugWarnings.
  • Updated phase-value reference table.
  • Updated the “Recording clinical advance data” section to use drugAndClinicalCandidates.count rather than knownDrugs count + filter.
  • Added a schema-note callout flagging the migration so a reader hitting an old query gets a pointer.

Discovered while running the SOP for the senolytic question — querying for compounds targeting EPHA2/EPHB4/EFNB1/PIK3CD/SRC as dasatinib alternatives. The corrected queries surfaced vandetanib (multi-EPH+SRC), leniolisib (selective PI3Kδ, non-oncology approval), and 6 approved PI3Kδ inhibitors as the candidate landscape; that synthesis was conversational, not seeded into a wiki page.

[2026-05-05] verify — molecules/proteins/ku70-ku80.md

Pages verified: 1

  • molecules/proteins/ku70-ku80.md — 12 corrections; verified: true (partial scope — Seluanov 2007 PDF not retrievable via a local paper archive [green OA, PMC:PMC2699370, download failed]; Gu 1997 PNAS not downloaded; canonical-DB identity fields not re-checked)

Sources checked:

  • doi:10.1038/35088000 (Walker 2001 Nature) — PDF verified end-to-end; crystal structure paper
  • doi:10.1038/382551a0 (Nussenzweig 1996 Nature) — PDF verified end-to-end; Ku80-null mice
  • doi:10.1016/s1074-7613(00)80386-6 (Gu 1997 Immunity) — downloaded 2026-05-05 and verified end-to-end; Ku70-null mice
  • doi:10.1016/j.dnarep.2007.06.010 (Seluanov 2007 DNA Repair) — download failed; PMC:PMC2699370; no-fulltext-access applied to Seluanov-derived section
  • doi:10.1073/pnas.94.15.8076 (Gu 1997 PNAS) — not downloaded; pending

Corrections:

  1. Central channel width “~22 Å” → “radius ~11.5 Å (diameter ~23 Å)” per Walker 2001’s Van der Waals exclusion volume calculation (the “22 Å” figure does not appear in Walker 2001)
  2. Ku70/Ku80 Ku-domain residue ranges “261–468 / 253–452” → “central polypeptide rings: Ku70 277–341 / Ku80 267–336” with N-terminal α/β domains Ku70 34–250 / Ku80 6–238; the previous ranges were not from Walker 2001
  3. VWFA domain attribution for Ku80 N-terminal domain — Walker 2001 calls it a Rossman fold/α/β domain, not VWFA; tagged unsourced pending primary citation for VWFA classification
  4. EEXXXDDL motif residues 720–728 — not stated in Walker 2001 (Walker mentions only the ~19 kDa C-terminal domain); tagged unsourced
  5. Ku70 SAP domain “residues 573–607” — Walker 2001 places the SAP domain start at ~536 with disordered linker 539–558; the 573–607 range is not directly stated; tagged unsourced
  6. Kd “~1 nM” → “1.5–4.0 × 10⁻¹⁰ M (~0.15–0.4 nM)” per Walker 2001 citing Yoo & Dynan 1999 (ref. 22); original figure was already tagged unsourced but the actual value was wrong by 3–7x
  7. Nussenzweig 1996 footnote strain “C57BL/6” → “mixed 129/Sv × C57BL/6J background, injected into C57BL/6 blastocysts” per paper methods
  8. Gu 1997 footnote — updated from “pending” to locally available; full author list added; strain corrected to “mixed 129/Sv × C57BL/6J background”
  9. Ku70-null body weight “~50%” → “50–60% of littermate controls” per Figure 1B in Gu 1997; Ku80-null weight now specified separately as “40–60%” per Nussenzweig 1996
  10. Lymphomagenesis claim for Ku80-null mice corrected — Nussenzweig 1996 explicitly states Ku80-null mice are NOT predisposed to early thymic tumors; removed incorrect “similar tumor predisposition” claim; added specific Ku70 tumor data (6 cases in ~40 mice, ages 2–7 months, CD4+CD8+TCRβ- phenotype) from Gu 1997
  11. Progeroid tissue changes (skin atrophy, osteopenia, liver dysfunction) — not reported in Gu 1997 or Nussenzweig 1996; replaced with what IS reported (premature MEF senescence in Ku70-null); tagged unsourced for tissue-level claims
  12. T cell phenotype nuance added — Ku70-null shows “leaky” T cell development (some DP cells present) per Gu 1997; distinct from complete T cell block in Ku80-null per Nussenzweig 1996
  13. Seluanov 2007-derived section flagged with unverified banner; PMC accession added to footnote; download failure documented

Downstream pages needing update:

  • hallmarks/genomic-instability.md — if it cites Ku70/80 mouse phenotypes including the incorrect skin atrophy/osteopenia/liver claims, those need correction
  • processes/non-homologous-end-joining.md (stub, not yet created) — when created, structural data from Walker 2001 should match corrected residue ranges

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

Pages verified: 1

  • pathways/nucleotide-excision-repair.md — 9 corrections; verified: true (partial scope — Marteijn 2014 and de Boer 2002 are not_oa and remain unverified against PDF; Niedernhofer 2006 and Vermeij 2016 verified against local PDFs)

Sources checked:

  • doi:10.1038/nature05456 (Niedernhofer 2006 Nature) — PDF verified end-to-end; local path confirmed
  • doi:10.1038/nature19329 (Vermeij 2016 Nature) — PDF downloaded during this pass and verified end-to-end
  • doi:10.1038/nrm3822 (Marteijn 2014 Nat Rev Mol Cell Biol) — not_oa; no local PDF; mechanistic NER steps remain unverified against PDF
  • doi:10.1126/science.1070174 (de Boer 2002 Science) — not_oa; no local PDF; TTD mouse lifespan/phenotype claims unverified against PDF

Corrections (Niedernhofer 2006): (1) Allele genotype Ercc1^-/Δ^ → Ercc1^-/-^ pure null; (2) n footnote “not stated” → n=27 (Fig. 2b caption); (3) Lifespan “median ~4 weeks” → “typically by 4 weeks; median ~3 weeks from KM curve”; (4) Progeroid feature list revised: Purkinje cell claim removed (not in Niedernhofer 2006 main text), replaced with paper-attested features; (5) Transcriptome tissue scope “liver, kidney, and brain” → “liver” (primary; Spearman r=0.32, P≤0.0001); kidney by qRT-PCR only; brain not in main-text comparison; (6) Footnote background “C57BL/6” → “not specified in main text”; (7) Cross-link corrected to distinguish Ercc1^-/-^ and Ercc1^Δ/-^ (F1 hybrid) backgrounds.

Corrections (Vermeij 2016 — new section added): (8) Section and footnote added; lifespan extension ~250% males / ~200% females (primary cohort; p<0.0001); ~180% replication cohort; F1 C57BL6J/FVB hybrid background; ~50% neuron retention; γH2AX reduction; Xpg^-/-^ ~80% extension noted.

Downstream pages needing update:

  • molecules/proteins/ercc1.md — Vermeij 2016 footnote lists “pending” (now downloaded); hedged ~200% should be updated to sex-specific figures and F1 hybrid background; Niedernhofer 2006 section should clarify Ercc1^-/-^ vs Ercc1^Δ/-^ distinction

[2026-05-05] verify — molecules/proteins/parp1.md

Pages verified: 1

  • molecules/proteins/parp1.md — 7 corrections; verified: true (partial scope — Bürkle 2005 closed-access, not verified; canonical-DB identity fields not re-checked)

Sources checked:

  • doi:10.1038/nature03443 (Bryant 2005 Nature) — PDF verified end-to-end
  • doi:10.1038/nature03445 (Farmer 2005 Nature) — PDF verified end-to-end
  • doi:10.1056/NEJMoa0900212 (Fong 2009 NEJM) — PDF verified end-to-end
  • doi:10.1016/j.cell.2013.06.016 (Mouchiroud 2013 Cell) — PDF downloaded and verified end-to-end
  • doi:10.1016/j.biocel.2004.10.006 (Bürkle 2005 Int J Biochem Cell Biol) — closed-access (not_oa); no-fulltext-access retained

Corrections:

  1. Fong 2009 — trial population: “60 patients with advanced solid tumours carrying BRCA1/2 mutations” → “60 total patients (22 confirmed BRCA1/2 carriers; remainder wild-type or unknown BRCA status)” — the trial enrolled all-comers; BRCA mutation was NOT an enrollment requirement
  2. Fong 2009 — “63% objective response rate in BRCA-mutated ovarian cancer patients (12/19)” → 12/19 (63%) is the clinical benefit rate (response + stable disease ≥4 months), not ORR; 9/19 (47%) is the RECIST objective response rate in BRCA carriers with ovarian/breast/prostate cancer
  3. Fong 2009 — “47% response rate in BRCA-mutated breast cancer (9/19)” → removed; 9/19 is the overall BRCA-carrier radiologic/tumor-marker response rate, not a breast-cancer-specific figure; breast cancer arm had 3 BRCA2-mutated patients (1 complete remission, 1 stable disease); full corrected breakdown added
  4. Mouchiroud 2013 — “PARP-2 deletion” → corrected to “pme-1 mutation (worm PARP-1 homolog) and pharmacological PARP inhibition”; pme-1 is the worm PARP-1 homolog, not PARP-2; the dominant worm PARP activity
  5. Mouchiroud 2013 — “NMN, NR” as NAD+ precursors → corrected to “NR and NAM”; NMN does not appear in Mouchiroud 2013 (NMN is from Yoshino 2011); Mouchiroud uses NR and NAM in worms, NR in mammalian cells
  6. Bryant 2005 footnote — “BRCA2-deficient DT40” → corrected to “BRCA2-deficient V-C8 (Chinese hamster) and human breast cancer cell lines”; DT40 is a chicken cell line used in other labs; Bryant uses V-C8, MCF-7, MDA-MB-231, SW480SN.3; xenograft n=40 CD-1 nude mice added
  7. Farmer 2005 footnote — “BRCA1/2-deficient human cancer cell lines” → corrected to “mouse embryonic stem cells (primary model)”; Farmer’s main experiments use isogenic BRCA1-deficient (Cre10) and BRCA2-deficient (Cre24) mouse ES cells; human MCF7 cells are supplementary only; in vivo model is teratoma xenograft in BALB/c-nude mice; p-values corrected to P<0.05 (siRNA) and P=0.03/P=0.01 (in vivo)

Unverifiable claims:

  • Bürkle cross-species correlation (r≈0.8, P<0.01, 13 species): closed-access PDF; numbers retained from original extraction with no-fulltext-access
  • Canonical-database identity fields (UniProt P09874, NCBI Gene 142, HGNC 270, Ensembl ENSG00000143799, GenAge 60): not re-checked against source databases in this pass

Downstream pages to check (main agent):

  • studies/fong-2009-olaparib-brca-phase1.md — study page likely carries the same wrong ORR/CBR confusion and wrong population description; propagate corrections
  • studies/mouchiroud-2013-nad-sirtuin-longevity.md — study page may carry “PARP-2 deletion” and “NMN” errors; propagate pme-1 and NR/NAM corrections
  • molecules/proteins/sirt1.md — verified (Mouchiroud cross-reference); check whether sirt1.md carries the “PARP-2 deletion” or “NMN” framing and correct if so
  • molecules/compounds/ (any olaparib or PARP inhibitor page) — may carry the 63% ORR label or wrong trial population; check

[2026-05-05] verify — pathways/homologous-recombination.md

Pages verified: 1

  • pathways/homologous-recombination.md — 6 corrections across 4 primary-source PDFs; verified: true (partial scope — White 2016, Callen 2020, San Filippo 2008, Moynahan 2001-BRCA2, Lord 2015 not PDF-verified)

Sources checked:

  • doi:10.1038/nature03443 (Bryant 2005 Nature) — PDF verified end-to-end; locally downloaded
  • doi:10.1038/nature03445 (Farmer 2005 Nature) — PDF verified end-to-end; locally downloaded
  • doi:10.1056/NEJMoa0900212 (Fong 2009 NEJM) — PDF verified end-to-end; locally downloaded
  • doi:10.1016/s1097-2765(00)80202-6 (Moynahan 1999 Mol Cell) — PDF downloaded and verified end-to-end
  • Reactome IDs R-HSA-5685942 (HR) and R-HSA-5693571 (NHEJ) confirmed via Reactome ContentService API — frontmatter R-HSA-5685942 is correct

Corrections:

  1. Moynahan 1999 body claim — “~5-fold reduction” → “5- to 6-fold reduction” (paper reports consistent range across two loci)
  2. Moynahan 1999 body claim — “NHEJ is unaffected” → “no gross NHEJ defect observed — NHEJ modestly elevated 1.5- to 1.6-fold in Brca1-null cells” (paper explicitly reports the elevation; “no gross defect” is paper’s own phrasing)
  3. Moynahan 1999 footnote — fixed HR fold-reduction to “5- to 6-fold”; corrected “NHEJ unaffected” to accurate description; added full author list; removed stale “archive: pending” status
  4. Bryant 2005 / Farmer 2005 body text — “PARP inhibitors selectively killed BRCA2-deficient cells at concentrations 1000-fold lower… [^bryant2005]” → attribution split correctly: Bryant demonstrated selective killing with NU1025/AG14361 (xenograft + cell lines); Farmer demonstrated quantitative fold-selectivity (57-fold BRCA1, 133-fold BRCA2 in ES cells; >1000-fold in CHO cells) with KU0058684/KU0058948
  5. Bryant 2005 footnote — removed incorrectly attributed KU0058948 compound (Farmer’s compound); replaced with correct compounds NU1025 and AG14361; added accurate xenograft n (40 mice, 10/20 V-C8 and 9/20 V-C8+B2 take rate); added accurate efficacy result (3/5 V-C8 xenografts responded)
  6. Fong 2009 body claim — “objective tumor responses in 12/19” → “clinical benefit in 12/19 (63%); 9/19 objective radiologic response by RECIST” — paper uses two distinct endpoints; prior text conflated them
  7. Fong 2009 footnote — added endpoint distinction (clinical benefit vs RECIST response); added maximum administered dose (600 mg BID); enumerated dose-limiting toxicities

Unverifiable claims:

  • White 2016 (BRCA1 expression decline with age in tissues): archive status pending; claims tagged [^white2016] unverified
  • Callen 2020 (53BP1 dual-block mechanism): archive status pending; claims tagged [^callen2020] unverified
  • San Filippo 2008 (mechanistic review): not_oa; multiple mechanistic claims sourced here unverified; no-fulltext-access in footnote
  • Moynahan 2001 BRCA2 (RAD51 focus abrogation): archive pending; claims tagged [^moynahan2001brca2] unverified
  • Lord 2015 (PARP inhibitor resistance mechanisms): not_oa; no-fulltext-access in footnote

Downstream pages to check:

  • molecules/proteins/brca1.md (R19 batch) — cites this pathway page and may carry the 5-fold Moynahan claim; NHEJ elevation finding may require update on that page

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

Pages verified: 1

  • molecules/proteins/shelterin.md — 7 corrections; verified: true (partial scope — de Lange 2005, de Lange 2009, Liu 2004 Nat Cell Biol, Hockemeyer 2007 not PDF-verified; UniProt lengths all confirmed)

Sources checked:

  • doi:10.1038/ncb1275 (Celli & de Lange 2005 Nat Cell Biol) — PDF verified end-to-end (actual path: (local PDF); CLI reports stale (stale local path) path — FEATURE REQUEST implied)
  • doi:10.1038/ncomms1708 (Hewitt 2012 Nat Commun) — PDF downloaded and verified end-to-end
  • doi:10.1016/j.cell.2006.04.044 (Hockemeyer 2006 Cell) — PDF downloaded and verified end-to-end
  • UniProt P54274, Q15554, Q9NUX5, Q9BSI4, Q96AP0, Q9NYB0 — all six subunit lengths confirmed via REST API 2026-05-05
  • doi:10.1101/gad.1346005 (de Lange 2005 Genes Dev) — diamond OA but HTTP 520 on retry; PDF unavailable; no-fulltext-access tagged
  • doi:10.1126/science.1170633 (de Lange 2009 Science) — no OA candidate URLs; PDF unavailable; no-fulltext-access tagged
  • doi:10.1038/ncb1142 (Liu 2004 Nat Cell Biol) — closed-access (not_oa); no-fulltext-access tagged
  • doi:10.1038/nsmb1270 (Hockemeyer 2007 Nat Struct Mol Biol) — closed-access (not_oa); no-fulltext-access tagged

Corrections:

  1. Celli 2005 section — “Artemis, APE1” as the processing factors tested → corrected to “DNA ligase IV (NHEJ)” — the paper tests Lig4-/- cells and ERCC1/XPF context; Artemis and APE1 are not tested
  2. Celli 2005 section — added quantitative detail: fusions ~50-fold more frequent; 3’ overhang signal 119 ± 18% retained in Lig4-/- p53-/- TRF2-/- cells; mixed 129/BL6 background; n=5 embryos + 12 clonal lines; ~120-fold reduction in fusions in Lig4-/- background
  3. Hewitt 2012 section — “tissues from aged humans and mice” → “human MRC5 fibroblasts (in vitro) and gut/liver of C57BL/6 mice (ages 12–42 months)” — paper contains NO aged human tissue data
  4. Hewitt 2012 section — SASP via “NF-κB and cGAS-STING” falsely attributed to [^hewitt2012] → attribution removed; paper does not discuss NF-κB or cGAS-STING; correct attribution noted to Rodier 2009 / post-2012 cGAS-STING literature
  5. Hewitt 2012 section — added quantitative correlation coefficients from paper: liver TAF R=0.98, P=0.01; gut TAF R=0.95, P=0.02; n=3 mice per age group; n=15 for in vitro experiments
  6. POT1 section (Hockemeyer 2006) — “ATR activation, CHK1 phosphorylation, and fragile telomere phenotypes” → corrected to actual paper findings: TIF phenotype (γH2AX/53BP1 foci), endoreduplication (~17% metaphase spreads), DKO TIF in 70–80% nuclei; CHK1 not demonstrated in this paper; noted as inferred
  7. Mouse/human section — “TERT-KO late-generation mice [^hockemeyer2006]” → wrong citation; Hockemeyer 2006 does not describe this model; removed citation, tagged unsourced, noted Blasco lab work as likely source
  8. Extrapolation table for Hewitt 2012 updated: “aged human tissues confirmed” → partial (in vitro human cells confirmed; in vivo data from mice only)

Unverifiable claims remaining:

  • De Lange 2005 architectural details (TRFH domain residue ranges, tankyrase interaction) — not PDF-verified
  • De Lange 2009 t-loop mechanism claims — not PDF-verified
  • Hockemeyer 2007 TPP1-POT1 telomerase processivity enhancement — not PDF-verified
  • Liu 2004 PTOP/TPP1 identification — not PDF-verified
  • DC/cancer syndrome genetics (TINF2, ACD, POT1 mutations) — secondary sources; no primary papers cited

Downstream pages to check (main agent):

  • Any page citing hewitt-2012-telomere-ddr-senescence for “aged human tissue” or SASP/NF-κB/cGAS-STING claims — correct if present
  • hallmarks/telomere-attrition.md — likely cites shelterin page claims; verify TAF description is now mouse-tissue-only for in vivo aging data
  • processes/cellular-senescence.md — may cite TIF-to-SASP mechanism; verify SASP attribution is not credited solely to Hewitt 2012

Schema escalation noted:

  • The causes: frontmatter field lists downstream effects of shelterin LOSS-of-function (“dna-damage-response”, “cellular-senescence”, “sasp”) rather than effects of shelterin presence. This is semantically inverted relative to the field’s intent in entity pages for active proteins. Convention recommendation: for protective complexes, consider a loss-of-function-causes: field or document in the field comment that causes: on shelterin represents loss-of-function consequences. Escalation for main agent / user decision.

Archive tooling issue noted:

  • returns (stale local path) (stale path from a prior user config); actual file is at. Added to FEATURE_REQUESTS if not already present.

[2026-05-05] verify — molecules/proteins/gdf15.md

Pages verified: 1

  • molecules/proteins/gdf15.md — 3 corrections; verified: true (partial scope — four 2017 GFRAL papers closed-access; Wiklund 2010 download failed; Yatsuga 2015 substitution DOI-confirmed but not PDF-verified)

Sources checked:

  • doi:10.1073/pnas.94.21.11514 (Bootcov 1997 PNAS) — downloaded during session (green OA via PMC23523); PDF verified end-to-end; structural claims confirmed
  • doi:10.1038/s41586-019-1911-y (Coll 2020 Nature) — local PDF verified end-to-end; tissue-of-origin corrected
  • doi:10.18632/aging.101684 (Lu 2019 Aging) — local PDF verified end-to-end; GrimAge component list, n=6,935, HR=1.10, meta-P=2.0E-75 confirmed
  • UniProt Q99988 — live API confirmed mature chain aa 195–308 = 114 aa; Arg194 cleavage site confirmed
  • GenAge entry 23886 (Gdf15, Mus musculus) — live web confirmed: pro-longevity, +43% lifespan; manipulation type not shown in public table (gap remains)
  • doi:10.1111/j.1474-9726.2010.00629.x (Wiklund 2010 Aging Cell) — download failed (bronze OA URL 403); DOI confirmed correct via Crossref; no-fulltext-access tagged
  • doi:10.1038/nm.4392, nm.4393, nm.4394, nature24042 (four 2017 GFRAL papers) — all closed-access (not_oa); author-label assignments confirmed via archive title metadata; no-fulltext-access tagged

Corrections:

  1. Coll 2020 tissue of origin: “predominantly from intestinal epithelial cells” → “enterocytes of the distal small intestine and colon, and in periglomerular renal tubular cells” — paper’s mRNA tissue panel and in situ hybridisation show distal intestine + colon + kidney as primary sites; liver capable but not dominant in vivo source under oral dosing
  2. suomalainen2011 citation replaced: “Suomalainen 2011” paper does not exist as described; the Suomalainen lab established FGF21 (not GDF15) as a mitochondrial biomarker (Lehtonen 2016, Neurology); correct primary GDF15 paper is Yatsuga/Koga 2015 (Ann Neurol, doi:10.1002/ana.24506); footnote replaced and all body citations updated
  3. Bootcov 1997 footnote: “archive status: pending download” → “local PDF available (PMC23523)”; full journal citation details added

Unverifiable claims remaining:

  • Wiklund 2010 quantitative claims (HR, cohort n, effect size) — not verified; no-fulltext-access tagged on footnote
  • Four 2017 GFRAL papers — closed-access; GFRAL/RET signaling mechanism claims unverified against full text
  • GenAge 23886 manipulation type (KO vs OE) — not shown in public search table; gap noted in Limitations section
  • Yatsuga 2015 claims — DOI confirmed via PubMed; not verified against PDF (paper not in archive)

Downstream pages to check (main agent):

  • molecules/compounds/metformin.md — cites Coll 2020 tissue-of-origin; check if “intestinal epithelial cells” claim is present and needs correction to “distal intestine, colon, and kidney”
  • studies/coll-2019-gdf15-metformin.md — if this study page exists, tissue-of-origin claim may need same correction

[2026-05-05] verify — molecules/proteins/gdf11.md

Pages verified: 1

  • molecules/proteins/gdf11.md — 9 corrections; verified: true (partial scope — Sinha 2014, Katsimpardi 2014 PDFs pending; McPherron 1999 closed-access; canonical DB identifiers not re-checked)

Sources checked:

  • doi:10.1016/j.cell.2013.04.015 (Loffredo 2013 Cell) — local PDF verified end-to-end
  • doi:10.1016/j.cmet.2015.05.010 (Egerman 2015 Cell Metabolism) — local PDF verified end-to-end
  • doi:10.1161/CIRCRESAHA.115.307527 (Smith 2015 Circ Res) — downloaded during session (bronze OA); PDF verified end-to-end
  • doi:10.1038/10320 (McPherron 1999 Nat Genet) — permanently not_oa; developmental-biology claims not verifiable from full text; DOI confirmed correct (archive title matches expected paper)
  • doi:10.1126/science.1251152 (Sinha 2014 Science) — OA/green; download still pending at verification time; claims tagged no-fulltext-access
  • doi:10.1126/science.1251141 (Katsimpardi 2014 Science) — OA/bronze; download still pending at verification time; claims tagged no-fulltext-access

Corrections:

  1. “89% sequence identity” (GDF11 vs myostatin mature domain) → 90% — Egerman 2015 explicitly states “90% homology in the mature active regions”; corrected in intro, protein structure section, and key interactors section
  2. Loffredo 2013 assay method: “commercially available ELISA” → SOMAscan broad-scale aptamer-based proteomics (1,001 proteins; 20 mice) for discovery; Abcam anti-GDF11 western blot for confirmation — “commercial ELISA” is wrong; no ELISA was used
  3. Loffredo 2013 n: “n=8–12/group” → parabiosis groups n=4–12/group; rGDF11 injection study n=16 GDF11 / n=13 saline (primary experiment, Fig 7D); n=18/22 in the CD45.2-only replication
  4. Egerman 2015 assay method: “modified Luminex aptamer-based assay (SOMAscan)” → custom GDF11-specific immunoassay (MULTI-ARRAY plates, anti-human GDF11 clone 743833; validated not to detect myostatin) — SOMAscan was used by Loffredo (to show non-specificity), not by Egerman for the specific measurement
  5. Egerman 2015 result precision: GDF11 “increases with age” → “~1.4-fold increase in rat serum, p=0.0534 (trend, not significant at p<0.05)”; human trend also non-significant; GDF11 was below detection in mice
  6. Smith 2015 institution: “Washington University, St. Louis” → Temple University School of Medicine, Philadelphia (with Boehringer Ingelheim co-authors)
  7. Smith 2015 assay method: “mass spectrometry-based protein quantification” → antibody-based immunoassay (R&D Systems anti-GDF11; validated not to detect myostatin); NO mass spectrometry was used — this was a fabricated claim
  8. Smith 2015 treatment duration: not previously noted → 28 days (not 30); n=21 GDF11, n=22 vehicle explicitly stated from paper
  9. Gaps section: “mass-spectrometry data (Smith 2015) is most specific” → removed/corrected throughout; neither replication study used mass spectrometry

Pages unverifiable (no PDF):

  • McPherron 1999 — not_oa permanently; structural/developmental claims tagged (closed-access noted in footnote)
  • Sinha 2014 — download pending; claims tagged no-fulltext-access
  • Katsimpardi 2014 — download pending; claims tagged no-fulltext-access

Downstream pages to check (main agent):

  • studies/loffredo-2013-gdf11-cardiac-rejuvenation.md — may have inherited wrong assay type (“ELISA”), wrong n values; verify against gdf11.md corrections
  • studies/egerman-2015-gdf11-increases-with-age.md — may describe assay as “SOMAscan” or “Luminex aptamer”; needs correction to custom immunoassay
  • studies/smith-2015-gdf11-no-cardiac-rescue.md — may claim mass spectrometry; needs correction to R&D Systems antibody-based assay; institution needs correction to Temple University
  • hallmarks/altered-intercellular-communication.md — may cite GDF11 as declining with age or use 89% identity; check for stale claims

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

Pages verified: 1

  • molecules/proteins/tert.md — 5 corrections (see below); verified: true (partial scope)

Sources checked:

  • doi:10.1126/science.279.5349.349 (Bodnar 1998 Science) — local PDF verified end-to-end
  • doi:10.1002/emmm.201200245 (Bernardes de Jesus 2012 EMBO Mol Med) — local PDF verified end-to-end
  • doi:10.1016/j.cell.2008.09.034 (Tomás-Loba 2008 Cell) — downloaded during session (bronze OA via camoufox); PDF verified end-to-end
  • doi:10.1126/science.276.5312.561 (Lingner 1997 Science) — download attempted; infoscience mirror returns HTML only; status: failed; no-fulltext-access
  • doi:10.1126/science.1230062 (Horn 2013 Science) — permanently not_oa; C228T/C250T claims not verifiable from full text

Corrections:

  1. Bodnar 1998 cell types: “BJ skin fibroblasts” → “BJ foreskin fibroblasts” (paper’s consistent terminology)
  2. Bodnar 1998 passage count precision: vague “>20 additional population doublings” → “RPE ~20 PD beyond hTRT- mean; BJ 36 PD beyond hTRT- mean” (from results and Fig. 3)
  3. Bodnar 1998 nude-mouse claim removed: “did not form tumors in nude mice” is not in the paper; replaced with correct evidence: no gross transformation markers (no contact-inhibition loss, no low-serum growth); noted that nude-mouse assay was not performed
  4. Tomás-Loba 2008 background description: “super-Ink4a/Arf or super-p19Arf mice” → paper uses Sp53/Sp16/SArf terminology; clarified that three tumor suppressors are overexpressed; 40% → 40.2% (exact figure from Fig. 5A); added sample sizes (n=27 vs n=68); added cancer-free subgroup figures (18%, 38%)
  5. Tomás-Loba 2008 footnote: background and Lingner 1997 footnote updated; archive status updated for Tomás-Loba (downloaded) and Lingner (failed)

Pages unverifiable (permanently):

Downstream pages to check (main agent):

  • hallmarks/telomere-attrition.md — Bernardes 2012 numbers already confirmed consistent (verified-partial as of 2026-05-04); no propagation needed for those numbers

[2026-05-05] verify — biomarkers/phenoage-2018.md

Pages verified: 1

  • biomarkers/phenoage-2018.md — 9 corrections (see below); verified: true

Sources checked:

  • doi:10.18632/aging.101414 (Levine 2018 Aging) — downloaded during session from PMC5940111; PDF verified end-to-end
  • doi:10.1038/s43587-022-00357-y (Waziry 2023 Nature Aging) — already downloaded; PDF verified (CALERIE-2)
  • doi:10.1038/s43587-024-00793-y (Bischoff-Ferrari 2025 Nature Aging) — already downloaded; PDF verified (DO-HEALTH Bio-Age)

Corrections applied:

  1. Stage 1 algorithm: “Klemera-Doubal method (KDM)” → Cox penalized regression → Gompertz mortality model (paper Methods section explicit)
  2. NHANES III n: “~9,000+” → 9,926 (paper text: “analytical sample included 9,926 adults”)
  3. Stage 2 training cohort: “InCHIANTI, FHS, BLSA, WHI, GS:SFHS (~3,600 individuals)” → InCHIANTI only (n=456); WHI×2, FHS, NAS, JHS are validation cohorts only
  4. Biomarker units corrected per Table 1: albumin g/dL → g/L; creatinine mg/dL → umol/L; glucose mg/dL → mmol/L
  5. Mortality HR: added distinction between clinical PhenoAge HR=1.09 (NHANES IV Table 2) vs DNAm PhenoAge Meta HR=1.045 [1.039–1.051] p=7.9E-47 (Figure 2 five-cohort meta)
  6. Waziry 2023 footnote: added precise effect sizes — PhenoAge d=−0.03/d=0.05 (both NS); GrimAge d=−0.04/d=0.05 (both NS); DunedinPACE 24-mo d=−0.25 [−0.41, −0.09] P<0.003; corrected randomized N=220 vs analysis n=197
  7. DO-HEALTH: removed unsourced; resolved with Bischoff-Ferrari 2025 (omega-3 alone: PhenoAge d=−0.16 [−0.02, −0.30]; all-3 combined: additive d=−0.24 to −0.32; GrimAge NS); added [^bischoffferrari2025] footnote
  8. First author affiliation: “Yale → UCLA” → “UCLA, Department of Human Genetics, at time of publication”
  9. Archive status banner removed; PDF path updated in Identity section

Downstream pages updated (propagation by verifier, per MOC correction exception):

  • frameworks/biological-age-measurement.md — PhenoAge row training target corrected from “KDM mortality composite” to “Cox penalized regression → Gompertz mortality score, NHANES III”; mortality HR column expanded to distinguish DNAm vs clinical

Downstream pages for main agent propagation:

  • interventions/lifestyle/caloric-restriction.md — may cite PhenoAge CALERIE-2 null; check footnote has precise d values
  • biomarkers/dunedinpace-2022.md — comparison claims vs PhenoAge; may need DO-HEALTH context added

[2026-05-05] verify — biomarkers/lehallier-proteomic-clock-2019.md

Pages verified: 1

  • biomarkers/lehallier-proteomic-clock-2019.md — 12 corrections; verified: true (partial scope on MAE)

Sources checked:

  • doi:10.1038/s41591-019-0673-2 (Lehallier 2019 Nat Med) — downloaded and verified end-to-end from PMC7062043; PMC author manuscript lacks main figures (Fig. 1g), so MAE ~2.9 yr not confirmable from text alone

Corrections applied:

  1. Feature selection method: “Elastic-net regression” → LASSO (glmnet alpha=1, 10-fold CV) — paper states alpha=1
  2. Cohort sub-n values: “LonGenity (n=~1,400) + INTERVAL (n=~1,500)” → INTERVAL n=3,301 (18–76y); LonGenity n=962 (1,030 enrolled minus 68 excluded, 61–95y); 4 independent cohorts n=171 not part of 4,263
  3. Training/test split: added explicit split (n=2,817 training / n=1,446 validation)
  4. r-value: “r=0.94” → r=0.93–0.97 (discovery r=0.97; 4 independent cohorts r=0.93)
  5. MAE ~2.9 yr: flagged as unconfirmed in PMC text; appears only in main Fig. 1g (absent from PMC author manuscript)
  6. Wave 1 biology: “musculoskeletal and cardiovascular biology” → ECM/structural pathway protein downregulation
  7. Wave 2 biology: corrected to “hormonal activity, binding functions, blood pathways; CVD proteins enriched”
  8. Wave 3 biology: corrected to “blood pathways and BMP signaling; MMP12 and CHRDL1 prominent”
  9. Sex differences: “~1,000 proteins” → 895/1,379 age-associated proteins also sex-associated (q<0.05)
  10. GDF11/VEGF-C therapeutic claims: removed as Lehallier 2019 findings; attributed to separate Wyss-Coray lab work
  11. Horvath MAE comparison (3.6 yr): flagged unsourced from this paper
  12. Banner and archive status: removed ⚠️ banner; updated archive status to downloaded

Downstream pages with potentially stale claims:

  • frameworks/biological-age-measurement.md — comparison table row shows “~2.9 yr” MAE for Lehallier; unconfirmed from manuscript text

[2026-05-05] verify — biomarkers/grimage-2019.md

Pages verified: 1

  • biomarkers/grimage-2019.md — 7 corrections (see below)

Sources verified (PDFs read end-to-end):

  • doi:10.18632/aging.101684 (Lu 2019, GrimAge v1) — green OA, downloaded from PMC6366976, 25 pages
  • doi:10.18632/aging.204434 (Lu 2022, GrimAge2) — open access, local PDF, 66 pages (key sections read)

Corrections applied:

  1. Plasma protein surrogate list entirely wrong — tPA listed as a component; ADM and TIMP-1 absent. Corrected to: ADM, B2M, Cystatin C, GDF-15, Leptin, PAI-1, TIMP-1, packYears. tPA is not a GrimAge component.
  2. Validation cohorts entirely wrong — “ARIC, WHI, MRCA, InCHIANTI, NSHD, ESTHER” replaced with actual cohorts: FHS test (n=625), WHI BA23 (n=2,107), WHI EMPC (n=1,972), JHS (n=1,747), InCHIANTI (n=924 from 484 individuals); 6,935 individuals / 7,375 arrays across 5 cohorts.
  3. Training n corrected — “~1,651” → 1,731 (FHS 70% split from 2,356 total).
  4. HR framing corrected — vague “1.5–2.5 per SD” replaced with sourced HR=1.10/yr (95% CI 1.09–1.17; meta P=2.0E-75; Cochran Q P=0.16).
  5. n-cpgs-or-features filled — null → 1,030.
  6. Never-smoker caveat corrected — GrimAge remains significant in never-smokers (N=3,988, P=1.1E-16), not reduced as previously implied.
  7. GrimAge2 training n and validation scope added — n=1,833 training; 13,399 samples / 10,065 individuals / 9 cohorts validation; HR=1.10/yr (P=3.6E-167) vs v1 P=2.0E-144.

Downstream pages to check (main agent action):

  • biomarkers/phenoage-2018.md — may reference GrimAge validation cohort names erroneously
  • hallmarks/epigenetic-alterations.md — references GrimAge; check for tPA or wrong cohort claims
  • hypotheses/information-theory-of-aging.md — references GrimAge validation cohorts
  • interventions/lifestyle/caloric-restriction.md — GrimAge CALERIE-2 null confirmed correct; no action needed

[2026-05-05] verify — biomarkers/hannum-clock-2013.md

Pages verified: 1

  • biomarkers/hannum-clock-2013.md — 7 corrections; PDF downloaded during this pass (bronze OA via Cell.com)

Sources verified: doi:10.1016/j.molcel.2012.10.016 (Hannum 2013, Mol Cell 49:359–367) — read end-to-end

Corrections applied:

  1. Sex effect corrected: “~4 yr faster” → “~4% faster (AMAR ratio); F test p=6×10⁻⁶” — ratio not raw years
  2. Drift rate “0.04 yr/yr at older ages” removed — not in paper; tagged unsourced
  3. Centenarian-offspring “~2–3 yr younger” removed — no centenarian-offspring cohort in paper; replaced with 2-meQTL AMAR finding; tagged unsourced
  4. Validation cohort corrected: “Held out 30%” → “n=174 independent Hispanic cohort; R=0.905, RMSE=4.89 yr”
  5. CpG pool count corrected: “~71,000” → “70,387” (FDR<0.05 by F test)
  6. Training accuracy updated: “r=0.96” → “r=0.963, RMSE=3.88 yr” (Fig 2B); algorithm detail added
  7. Mortality prediction claim corrected: GrimAge comparison removed; Hannum 2013 did not report mortality outcomes; tagged unsourced

Downstream pages to check (main agent):

  • frameworks/biological-age-measurement.md — RMSE “~3–4 yr” captures training (3.88 yr) but not validation (4.89 yr); recommend precision update
  • hallmarks/epigenetic-alterations.md — no corrected claims inherited (generation classification + CALERIE-2 null only)
  • hypotheses/information-theory-of-aging.md — Hannum refs via Schooling 2025 MR only; no corrections needed

[2026-05-05] verify — biomarkers/dunedinpace-2022.md

Pages verified: 1

  • biomarkers/dunedinpace-2022.md — 5 classes of corrections (see below)

Sources verified (PDFs read end-to-end):

  • doi:10.7554/eLife.73420 (Belsky 2022) — gold OA; downloaded from PMC during this session
  • doi:10.1038/s43587-022-00357-y (Waziry 2023) — hybrid OA; local PDF already available

Corrections made:

  1. Training cohort n: “n=954” → “N=1037 total cohort; n=817 analysis subsample” (frontmatter, Identity section, Training Details table, and footnote)
  2. Test-retest ICC: “~0.90” → “0.96 (95% CI 0.93–0.98; Lehne replicate dataset)” (frontmatter, Training Details table, Performance Characteristics section, and footnote); second EPIC-array dataset ICC 0.97 (95% CI 0.94–0.98) added to ICC paragraph
  3. CALERIE 12-month effect: missing → added “12-month d=−0.29 (95% CI −0.45 to −0.13)” to body, summary table, and waziry2023 footnote
  4. Pre-registration status: “pre-registered primary endpoint” → nuanced: Waziry 2023 describes this as a “post hoc analysis” of CALERIE (NCT00427193); DunedinPACE was a primary endpoint of the DNAm sub-study but that sub-study was not independently pre-registered; corrected throughout
  5. Archive status line: “download pending” → “local PDF available (gold OA)”
  6. Auto-extraction banner removed.

Downstream propagation needed (main agent task):

  • biomarkers/telomere-length-leukocyte.md line 86: “DunedinPACE (ICC ~0.90)” → “DunedinPACE (ICC 0.96)”
  • frameworks/biological-age-measurement.md line 55: ICC column “~0.90” → “0.96”
  • frameworks/biological-age-measurement.md line 93: “ICC (~0.90)” → “ICC (0.96)“

[2026-05-05] verify — biomarkers/horvath-clock-2013.md

Pages verified: 1

  • biomarkers/horvath-clock-2013.md — 8 corrections

Sources checked:

  • doi:10.1186/gb-2013-14-10-r115 (Horvath 2013) — verified end-to-end, local PDF
  • doi:10.1038/s43587-022-00357-y (Waziry 2023) — verified for all cited claims, local PDF
  • doi:10.18632/aging.202913 (Fitzgerald 2021) — verified end-to-end, local PDF
  • doi:10.1038/s43587-022-00248-2 (Higgins-Chen 2022) — PDF pending; cited only for qualitative claim; not verified

Corrections applied:

  1. Frontmatter training-cohort: “8,120 samples” → “7,844 non-cancer samples” (n=7,844 stated explicitly in Results; 8,120 appears nowhere in paper)
  2. CALERIE-2 framing: “~12% achieved caloric restriction” → “mean achieved ~12.5% CR — roughly half the prescribed dose” (implied only 12% of participants achieved CR; incorrect)
  3. CALERIE-2 nuance: noted that Waziry 2023 primary outcomes were PC PhenoAge/GrimAge; first-gen Horvath in supplementary (also null)
  4. Fitzgerald 2021 n: “43 healthy adult males” → “44 randomized, 38 analyzed” (CONSORT diagram)
  5. Fitzgerald 2021 “TREAT trial” name removed — not in paper
  6. Fitzgerald 2021 p-value: “p<0.001” → “p=0.018”; unsourced CI (−4.17 to −2.28) removed — not in paper
  7. Fitzgerald 2021 detail added: within-group 1.96 yr, p=0.066 NS; saliva tissue; Metagenics COI; age 50–72
  8. DunedinPACE CI added: d=−0.25 (95% CI −0.41, −0.09) at 24 months

Downstream pages for main agent to check:

  • interventions/lifestyle/caloric-restriction.md — “~12% achieved” framing
  • hypotheses/information-theory-of-aging.md — CALERIE-2 Horvath null specifics
  • studies/fitzgerald-2021-diet-lifestyle-epigenetic-clock — stub; n/p-value/COI if expanded

[2026-05-05] verify — biomarkers/frailty-index.md

Pages verified: 1

  • biomarkers/frailty-index.md — 6 corrections (see below)

Sources verified (PDFs read end-to-end):

  • doi:10.1100/tsw.2001.58 (Mitnitski 2001) — gold OA, local PDF
  • doi:10.1093/gerona/glt136 (Whitehead 2014) — bronze OA, local PDF

Corrections applied:

  1. HR ~1.04 per 0.01 FI removed as Mitnitski 2001 claim — paper uses gamma-distribution survival model (V=q·T), not a Cox HR; figure appears in later Rockwood/Mitnitski work; tagged unsourced pending Rockwood 2007 (closed-access)
  2. Mouse FI age correlation corrected — r=0.66 → r²=0.91 exponential fit (clinical FI); r=0.66 was from the 8-item performance-based index
  3. Mouse FI mortality prediction qualified — Whitehead 2014 explicitly disclaims this; one animal showed FI acceleration before death, but authors state larger study needed
  4. Deficit-list prevalence cutoffs removed — unsourced thresholds (>80% / <1%) not stated in Mitnitski 2001; replaced with what the paper actually shows (resampling robustness with 20–60 item subsets)
  5. Cohort n corrected — “~2,900+” → n=2,913 exact (mortality analysis n=1,468)
  6. CSHA FI item count corrected — “40-item FI” → “92-item FI” (per Mitnitski 2001 Appendix)

Downstream pages to check (main agent action):

  • frameworks/biological-age-measurement.md — comparison table row for FI HR may be stale; check for “~1.04 per 0.01 FI” entry and qualifier needed

[2026-05-05] verify — biomarkers/telomere-length-leukocyte.md

Pages verified: 1

  • biomarkers/telomere-length-leukocyte.md — 6 corrections (see below)

Corrections applied (Codd 2021 PDF verified):

  1. IPF MR finding removed — IPF was not an MR outcome in Codd 2021; had been incorrectly attributed there; tagged needs-replication for proper citation
  2. Locus count corrected: “>130 loci” → “138 genomic loci (108 new); 197 sentinel variants”
  3. Sample size / study description corrected: “n>472,000 across multiple GWAS meta-analyses” → “n=472,174 UK Biobank single cohort”
  4. Life-expectancy finding added: 2.47-yr shorter life expectancy (95% CI 1.99–2.96) for LTL >1 s.d. below mean at age 40 — was absent from page
  5. Cancer directionality finding added: longer LTL causally increases risk of prostate, melanoma, thyroid, kidney, sarcoma, MS — critical nuance absent from original
  6. All-cause mortality MR framing corrected: paper does not run a direct all-cause mortality MR; it uses life-expectancy modeling from cause-specific mortality; the “null for all-cause mortality” claim was both inaccurate and misrepresented

Pages unverifiable (closed-access):

  • Cawthon 2003 (doi:10.1016/S0140-6736(03)12384-7) — not_oa; top-line n=143 and HR values consistent with published abstract; full adjustment details unverified; no-fulltext-access retained

Downstream pages to check (main agent action):

  • hallmarks/telomere-attrition.md — biomarker section does not directly cite Codd 2021; no Codd 2021-derived errors found; no corrections needed there
  • frameworks/biological-age-measurement.md — may reference LTL MR findings; check for IPF attribution and “null all-cause” framing

[2026-05-05] Round 17 — Biomarker layer seeded (DONE)

Deliverables

New directory: biomarkers/ (created)

New biomarker pages (8):

  • biomarkers/horvath-clock-2013.md — 353-CpG pan-tissue clock. doi:10.1186/gb-2013-14-10-r115 (local PDF available). Key: null in CALERIE-2 CR; positive in Fitzgerald 2021 pilot (contested).
  • biomarkers/hannum-clock-2013.md — blood-specific 71-CpG clock. doi:10.1016/j.molcel.2012.10.016 (pending). Key: null in CALERIE-2; centenarian offspring ~2–3 yr younger.
  • biomarkers/phenoage-2018.md — 9-biomarker + 513-CpG mortality clock. doi:10.18632/aging.101414 (pending). Key: null in CALERIE-2; HR ~1.045/yr; MR null.
  • biomarkers/grimage-2019.md — plasma-protein-surrogate + smoking DNAm clock; strongest mortality HR. doi:10.18632/aging.101684 (pending). GrimAge2 doi:10.18632/aging.204434 (local PDF). Key: null in CALERIE-2; MR null; smoking confound noted.
  • biomarkers/dunedinpace-2022.md — 173-CpG pace clock; ICC ~0.90. doi:10.7554/eLife.73420 (pending). Key: CALERIE-2 POSITIVE (d=−0.25 p<0.003); only clock to respond to CR.
  • biomarkers/lehallier-proteomic-clock-2019.md — 373-protein SomaScan plasma clock. doi:10.1038/s41591-019-0673-2 (pending). Key: non-monotonic waves at 34/60/78 yr; no RCT data yet.
  • biomarkers/telomere-length-leukocyte.md — LTL biomarker page. doi:10.1038/s41588-021-00944-6 (Codd 2021 MR, local PDF); doi:10.1016/S0140-6736(03)12384-7 (Cawthon 2003, closed access). Key: MR null for all-cause mortality; causal for CAD and IPF; poor test-retest.
  • biomarkers/frailty-index.md — deficit accumulation FI. doi:10.1100/tsw.2001.58 (Mitnitski 2001, local PDF); doi:10.1093/gerona/glt136 (Whitehead 2014 mouse FI, local PDF). Key: HR ~1.04/0.01 FI; mouse validated; intervention-responsive (exercise).

New framework page (1):

  • frameworks/biological-age-measurement.md — MOC comparing all clock types; tradeoffs; reproducibility critique (Higgins-Chen 2022); intervention-responsiveness debate; external tool references (dnamage.clockfoundation.org, Open Aging Atlas).

Cross-page propagation:

  • interventions/lifestyle/caloric-restriction.md — added ## Biomarker effects section with DunedinPACE (positive, CALERIE-2) and Horvath/PhenoAge/GrimAge (null, CALERIE-2) with cross-links.
  • hypotheses/information-theory-of-aging.md — added ## Biological-age clock data section documenting which clocks support vs challenge the theory; cross-linked all 5 DNAm clock pages.
  • hallmarks/epigenetic-alterations.md — added all 5 clock pages and frameworks/biological-age-measurement to cross-references table.

DOI verification summary (see below in summary)

All 8 primary DOIs confirmed via. 5 have local PDFs; 3 are pending download.

Gaps surfaced

  • DO-HEALTH PhenoAge signal: #gap/unsourced — specific DO-HEALTH clock results need direct citation
  • Fitzgerald 2021 PhenoAge assessment: #gap/unsourced — unclear if PhenoAge was an endpoint
  • Exercise-GrimAge RCT evidence: #gap/unsourced
  • Mouse FI rapamycin/acarbose data: #gap/unsourced
  • DunedinPACE MR (causal validation): not-tested — no published MR instrument as of 2026-05-05

Graph View

Backlinks