🧬 Aging Wiki

R33

28 min read

log/R33.md — Round 33 entries

Sub-file of log — see parent for index.

[2026-05-07] verify | pathways/telomerase-pathway.md

Pages verified: 1

  • pathways/telomerase-pathway.md — corrections: 5 factual changes; 2 new footnotes added; 2 seeder PMIDs corrected to proper values

PDFs read and verified end-to-end:

  • Bodnar 1998 (10.1126/science.279.5349.349) — local PDF (pre-existing in archive)
  • Bernardes de Jesus 2012 (10.1002/emmm.201200245) — local PDF (pre-existing in archive, gold OA)
  • Nguyen 2018 (10.1038/s41586-018-0062-x) — downloaded this pass via PMC (PMC6223129)
  • Vogan 2016 (10.7554/eLife.18221) — downloaded this pass via PMC (PMC5005035)

Sources not verified (permanently not_oa or download-failed):

  • Greider 1985 (10.1016/0092-8674(85)90170-9) — not_oa; no-fulltext-access retained
  • Cohen 2007 (10.1126/science.1138596) — not_oa; no-fulltext-access retained
  • Heaphy 2011 (10.1126/science.1207313) — not_oa; no-fulltext-access retained
  • Lingner 1997 (10.1126/science.276.5312.561) — download failed (infoscience mirror HTML only); no-fulltext-access (consistent with tert.md verified-scope)

Infrastructure checks:

  • Reactome R-HSA-157579 (“Telomere Maintenance”): confirmed live and correctly named
  • KEGG: no standalone hsa telomerase pathway; hsa04218 (Cellular senescence) confirmed as the embedding map — kegg:null + needs-canonical-id framing is correct

Corrections made:

  1. Cajal body / TCAB1 section — MATERIAL CORRECTION: Body claimed “TCAB1-deficient cells have reduced TERT recruitment to telomeres” and CB association as “rate-limiting checkpoint.” Vogan 2016 actually shows TCAB1 KO in HCT116 and hESC lines maintained stable telomere homeostasis — Cajal body localization is NOT required for telomere maintenance. Section rewritten to reflect actual findings; efficiency-mechanism framing retained as a hypothesis for low-TERT-level stem cells. needs-replication added.
  2. Nguyen 2018 resolution — QUANTITATIVE CORRECTION: “~7–9 Å” → “7.7 Å and 8.2 Å (catalytic core and H/ACA lobe respectively) + 10.2 Å overall holoenzyme.” Structural description expanded from primary source.
  3. Bernardes 2012 strain — PRECISION CORRECTION: “C57BL/6 mice” → “>95% C57BL/6 background mice” (paper Methods).
  4. Bernardes 2012 footnote expanded: added eGFP vs mTERT p-values (1-yr: p=0.02; 2-yr: p=0.05, Log Rank), viral dose (2×10^12 vg/mouse), catalytically-inactive mTERT-DN control finding.
  5. Bodnar 1998 footnote expanded: corrected single-PD framing to two-experiment breakdown (RPE at PD 37 MPSV-hTRT; BJ at PD 58 MPSV-hTRT or PD 44 pZeoSV-hTRT); p-values added (P<10^-24 RPE; P<10^-6 BJ); telomere elongation quantities added (+3.7 kbp RPE; +7.1 kbp BJ); “hTRT” nomenclature noted.
  6. TPP1 TEL patch citations — PMID CORRECTIONS: Seeder PMIDs 25030370 and 22037700 were completely wrong (mapped to a nursing paper and a parasite immunology paper). Correct papers identified via PubMed: Nandakumar 2012 (doi:10.1038/nature11648, PMID 23103865, Nature) and Schmidt 2014 (doi:10.7554/eLife.03563, PMID 25271372, eLife). Two new footnotes added. Neither PDF locally verified — both tagged no-fulltext-access.
  7. Auto-extraction banner removed; verified: true flipped; verified-scope populated.

Cross-check with verified tert page:

  • Bodnar 1998 and Bernardes 2012 numbers on tert.md match — consistent. No divergence.
  • tert.md’s Bernardes 2012 description uses the correct >95% C57BL/6 framing; telomerase-pathway.md now matches.

Downstream pages that may need updates (main agent to handle):

  • hallmarks/telomere-attrition.md — may use “C57BL/6” for Bernardes 2012 without the >95% qualifier; check
  • molecules/proteins/tpp1.md (if seeded) — likely inherits wrong PMIDs for TEL patch citations; check
  • Any page describing Cajal bodies as a required telomerase-recruitment checkpoint — Vogan 2016 finding overturns this framing
  • No reverse personal → research wikilinks introduced this pass; one-way discipline preserved.

[2026-05-07] verify | molecules/proteins/trf1.md

Pages verified: 1

  • molecules/proteins/trf1.md — corrections: 4 (see below)

PDFs / sources read:

  • van Steensel & de Lange 1997 (10.1038/385740a0) — local archive; PDF verified end-to-end
  • Smogorzewska et al. 2000 (10.1128/MCB.20.5.1659-1668.2000) — downloaded this pass (green OA via PMC); PDF verified end-to-end
  • Martínez et al. 2009 (10.1101/gad.543509) — downloaded this pass (diamond OA); PDF verified end-to-end
  • Sfeir et al. 2009 (10.1016/j.cell.2009.06.021) — local download failed (status: failed after retry); verified via PMC full text (PMC2723738)
  • Chong 1995 (10.1126/science.270.5242.1663) — not_oa; unverifiable
  • Smith 1998 (10.1126/science.282.5393.1484) — not in archive; unverifiable; referenced in context
  • UniProt P54274 — REST API verified (length 439 aa, TRFH 58–268, Myb 375–432, Ser219 phospho, SUMOylation Lys213/325/366 confirmed)
  • NCBI Gene 7013 — verified (TERF1, 8q21.11, Homo sapiens)
  • Open Targets ENSG00000147601 — queried 2026-05-07; tractability labels obtained; tier-3 aging-context confirmed

Corrections made:

  1. Martínez 2009 phenotype: removed fabricated “intestinal crypt failure” and “Lgr5+ intestinal stem cell depletion” claims — K5-Cre targets stratified epithelia only; no intestinal phenotype exists in this paper. Replaced with accurate description of the epidermal/hair follicle phenotype with quantitative survival data (8% survival at P3).
  2. Martínez 2009 genotype name corrected: “K5-Cre; Terf1^fl/fl” → “TRF1^Δ/Δ K5-Cre” (paper’s nomenclature).
  3. PARylation citation corrected: citation [^vansteensel1997] on the PARylation/tankyrase claim replaced with [^smith1998] — the 1997 paper established length regulation by protein counting; tankyrase/PARylation mechanism was established by Smith et al. 1998. New [^smith1998] footnote added with no-fulltext-access tag.
  4. Sfeir 2009 section expanded from qualitative description to quantitative: ~20% fragile telomeres TRF1-null vs <5% controls; aphidicolin raises to ~28%; SMARD replication fork stall: 7/97 vs 0/78 molecules; ATR/CHK1 (not ATM/CHK2) activated; BLM and RTEL1 helicases required.
  5. Open Targets gap tag updated: removed stale “API returned 400/500 errors” note; replaced with actual API result interpretation (2026-05-07 query successful; “Approved Drug” labels are indirect pathway associations via tankyrase; tier-3 aging-context confirmed).
  6. Chong 1995 footnote: added no-fulltext-access tag.
  7. Druggability limitations note: updated to reflect confirmed API query.

Pages unverifiable (closed-access):

  • Chong 1995 (10.1126/science.270.5242.1663) — tagged no-fulltext-access in footnote
  • Smith 1998 (10.1126/science.282.5393.1484) — not in archive; no-fulltext-access in footnote

Downstream pages that may need updates:

  • stem-cell-exhaustion — cross-references “Lgr5+ intestinal stem cells” and “intestinal crypt” in the general stem-cell section; these are attributed to general biology, NOT to Martínez 2009 specifically, so no correction needed there
  • shelterin — correctly describes TRF1 role; no corrections needed; already verified
  • No study pages for van Steensel 1997, Smogorzewska 2000, or Sfeir 2009 exist yet; if seeded, they should inherit the corrected quantitative data from this pass

[2026-05-07] verify — studies/horvath-2013-epigenetic-clock.md

Pages verified: 1

  • studies/horvath-2013-epigenetic-clock.md — corrections: 7 substantive (see below)

Sources checked:

  • Horvath S (2013) Genome Biology 14:R115 — doi:10.1186/gb-2013-14-10-r115 — local PDF read end-to-end (all 19 pages); gold OA; local path

Corrections made:

  1. Cancer sample n: “6,000 cancer samples across 20 cancer types” → “5,826 cancer samples from 32 cancer tissue datasets (covering 20 cancer types).” Abstract uses ~6,000 colloquially; body text p.10 gives n=5,826 precisely. 32 = dataset count; 20 = cancer-type count.
  2. Cancer age acceleration: “mean of 36 years” → “average of 36.2 years” (Additional file 13B per paper body p.10).
  3. Cerebellum deceleration claim REMOVED: Paper explicitly states no significant difference in DNAm age between cerebellum and occipital cortex from the same subjects (Additional file 5K,L). Higher test-set error (5.9 yr) for cerebellar samples reflects calibration noise, not a within-subject directional deceleration. Section replaced with accurate brain-region analysis description.
  4. HIV and Down syndrome acceleration claims REMOVED: Neither finding appears anywhere in the primary paper. These findings originate from later studies (HIV: Gross et al. 2016; Down syndrome: Horvath et al. 2015 Genome Med). Correction note added; unsourced applied for downstream pages.
  5. Heritability: qualitative “modest but non-zero” → specific values: 100% in newborns (data set 50) and 39% in older subjects (data set 41), via Falconer’s formula on MZ+DZ twin pairs (p.8).
  6. Key-findings frontmatter updated: removed “cerebellum-deceleration” and “HIV-acceleration”; corrected cancer acceleration to “36.2yr”; heritability tag updated with specific values; “progeria-werner-syndrome-age-acceleration” added (confirmed in paper).
  7. Results table expanded: training vs held-out test accuracy distinguished; cancer validation row corrected; tissue-level performance data added from Figure 2 captions.

Confirmed correct (no changes needed):

  • n=7,844 non-cancer samples; 353 CpGs; held-out test r=0.96, MAE=3.6 yr; iPSC/ESC near zero; 82 GEO datasets; 51 tissues; elastic-net alpha=0.5; log(age+1) transformation

Unverifiable claims:

  • HIV acceleration and Down syndrome acceleration — not in this paper; correct primary sources needed
  • Higgins-Chen 2022 reliability critique — cited qualitatively; PDF download pending per biomarker page

Downstream propagation needed (main agent tasks):

  • biomarkers/horvath-clock-2013.md (verified 2026-05-05) — CRITICAL: training-details table states “n=7,844 non-cancer + 276 cancer.” The 276 cancer figure is wrong; primary source gives 5,826 cancer samples. Main agent must correct and re-date.
  • Any wiki page citing Horvath 2013 for cerebellum deceleration — re-source or remove
  • Any wiki page citing Horvath 2013 for HIV or Down syndrome clock acceleration — re-source to correct primary papers (HIV: Gross et al. 2016; Down syndrome: Horvath et al. 2015)
  • hypotheses/information-theory-of-aging.md — check whether it attributes HIV/Down syndrome to Horvath 2013
  • frameworks/biological-age-measurement — orphan framework MOC; flagged for R34 housekeeping wiring pass

Final verified state: verified: true — full PDF end-to-end verification completed.

[2026-05-07] verify — processes/dna-methylation

Page verified: processes/dna-methylation.md Sources checked: 5 footnotes

  • Heyn 2012 (doi:10.1073/pnas.1120658109) — downloaded, read end-to-end (6 pp. + full figures). VERIFIED.
  • Bocklandt 2011 (doi:10.1371/journal.pone.0014821) — downloaded, read end-to-end (5 pp. + figures). VERIFIED.
  • Bird 2002 (doi:10.1101/gad.947102) — diamond OA; download failed twice (CSHL server HTTP 520). UNVERIFIABLE from local PDF; Europe PMC reports not open access. Key claims (70–80% CpG methylation, CpG island promoter landscape) are foundational, widely replicated, and consistent with Heyn 2012 internal data, but Bird 2002 PDF verification deferred.
  • Smith & Meissner 2013 (doi:10.1038/nrg3354) — not_oa. UNVERIFIABLE.
  • Ehrlich 2002 (doi:10.1038/sj.onc.1205651) — not_oa. UNVERIFIABLE.

Corrections made (4):

  1. Heyn 2012 global methylation magnitude: “~20% reduction” → “~9% relative reduction (7.5 percentage-point absolute; 80.5% → 73.0%)” — The paper reports WGBS average methylation level of 80.5% (newborn) vs 73.0% (centenarian, Y103). The “~20% reduction” stated in the wiki table was incorrect; 7.5 pp out of 80.5% is a ~9.3% relative reduction, not 20%.
  2. Heyn 2012 sample sizes: “n=7 (4 centenarians, 3 newborns); array-based; confirmatory WGBS” → “WGBS: n=1 newborn + n=1 centenarian; array extension: n=19 newborns + n=19 nonagenarians” — The footnote’s “n=7 (4 centenarians, 3 newborns)” was entirely wrong. The primary WGBS was n=1+1; the 450K microarray validation extended to 19+19. The “n=7” figure does not appear in the paper.
  3. Bocklandt 2011 prediction model: “three CpG sites” → “two cytosines (EDARADD and NPTX2 promoters)” — The paper’s final leave-one-out regression model used just two cytosines; three genes (EDARADD, TOM1L1, NPTX2) were validated by pyrosequencing, but TOM1L1 did not contribute significantly to the regression and was not included in the final two-cytosine model. Both body prose and the footnote corrected.
  4. Wilson & Jones 1983 gap tag updated — Seeder correctly tagged this unsourced when the DOI was not found. DOI now confirmed via Crossref: 10.1126/science.6844925 (Wilson VL, Jones PA · Science 220:1055–1057 · “DNA Methylation Decreases in Aging But Not in Immortal Cells”). Paper is closed-access (not_oa). Gap tag updated to no-fulltext-access with DOI recorded in Limitations section for future reference.

Closed-access claims (unverifiable):

  • Smith & Meissner 2013: DNMT3A/3B/DNMT3L heterodimer complex description; UHRF1-DNMT1 maintenance mechanism. These are canonical textbook mechanisms; flagged no-fulltext-access in footnote.
  • Ehrlich 2002: Parallel between aging and cancer hypomethylation of repetitive elements and CpG island hypermethylation (CIMP connection). Flagged no-fulltext-access in footnote.

Unverifiable claims (Bird 2002 PDF not accessible):

  • “70–80% of CpGs methylated in vertebrate genomes” — consistent with standard field consensus and corroborated by Heyn 2012 (which reports 80.5% for newborn); pending Bird 2002 PDF availability.
  • CpG island promoter coverage (“~60% of human gene promoters”) — widely cited figure; not independently verified here.
  • MBD protein family (MeCP2, MBD1-4) mechanism via NuRD/Sin3A — unverified against Bird 2002 primary text.

Seeder judgment confirmed correct:

  • Omitting Wilson & Jones 1983 citation and tagging unsourced was the right call — the seeder could not locate the DOI. DOI is now confirmed; citation can be added to the global hypomethylation section if the main agent chooses to do so (it is closed-access and adds historical context only; Heyn 2012 is the stronger citation for the quantitative claim).

Downstream propagation needed (main agent):

  • biomarkers/horvath-clock-2013.md — references Bocklandt 2011 as precursor; verify the “three CpG sites” vs “two cytosines” language if present.
  • hallmarks/epigenetic-alterations.md — may cite the Heyn 2012 ~20% figure or the 4+3 sample-size framing; check.
  • Any page citing Heyn 2012 for the “centenarian” methylation reduction — verify magnitude language.

Final verified state: verified: true with scope (Heyn 2012 + Bocklandt 2011 verified; Bird 2002 download failed; Smith & Meissner 2013 + Ehrlich 2002 not_oa).

[2026-05-07] verify — processes/dna-demethylation

Page verified: processes/dna-demethylation.md Sources checked (7 footnotes):

  • Tahiliani 2009 (10.1126/science.1170116): OA via PMC2715015; PDF download failed (archive status: failed); verified via PMC full-text XML via eutils API. TET1 chemistry (2OG + Fe(II) dioxygenase requirement) confirmed. ES cell 5hmC data confirmed. No tissue distribution data in this paper.
  • Globisch 2010 (10.1371/journal.pone.0015367): gold OA; PDF downloaded and read end-to-end. 5hmC tissue distribution data corrected.
  • Jaiswal 2014 (10.1056/NEJMoa1408617): local PDF read; also cross-checked against previously verified clonal-hematopoiesis and hematopoietic-stem-cells pages. CHIP prevalence corrected.
  • Lu 2020 (10.1038/s41586-020-2975-4): cross-checked against verified study page (studies/lu-2020-osk-vision-restoration, verified R31). Extended Data Fig. references corrected.
  • Ito 2011 (10.1126/science.1210597): not_oa — cannot verify; tagged no-fulltext-access.
  • He 2011 (10.1126/science.1210944): not_oa — cannot verify; tagged no-fulltext-access.
  • Kohli & Zhang 2013 (10.1038/nature12750): PDF download failed despite PMC listing; unverified; tagged no-fulltext-access.

Corrections made (5):

  1. 5hmC brain abundance claim wrong: wiki stated “~1–2% of cytosines in cerebellum Purkinje cells” — this number does not appear in Globisch 2010 and the Purkinje cell claim derives from Kriaucionis & Heintz 2009 (not cited on this page). Corrected to the actual Globisch 2010 values: 0.3%–0.7% of dG (CNS tissues); non-CNS medium tissues 0.15%–0.17% of dG; low tissues 0.03%–0.06% of dG. Unit clarified (% of dG, not % of all cytosines).
  2. Brain 5hmC range in aging section: “~0.4–1.5% of all cytosines” corrected to “0.3%–0.7% of dG” per Globisch 2010; unit (% of dG vs % of all cytosines) explicitly noted.
  3. Jaiswal 2014 TET2 prevalence numbers wrong and framing wrong: wiki claimed “~2–3% at age 60+” and “~5% at 80+” for TET2 specifically. The paper reports overall CHIP prevalence (all driver genes) of 5.6% at 60–69y / 9.5% at 70–79y / 11.7% at 80–89y / 18.4% at ≥90y. Corrected both the numbers and the framing (these are overall CHIP, not TET2-specific).
  4. Lu 2020 Extended Data Fig. reference corrected: wiki said “Extended Data Fig. 7” for the TET1/TET2 knockdown mechanism. Verified study page confirms TET1/TET2 KD = Extended Data Fig. 6e–f; TDG KD = Extended Data Fig. 7a–d. Corrected both inline body text and footnote.
  5. Footnotes updated: Tahiliani 2009 footnote now accurately describes what the paper demonstrates (ES cell + in-vitro data; no tissue distribution). Globisch 2010 footnote now has the correct quantitative tissue values. Jaiswal 2014 footnote now has correct CHIP prevalence by age bracket. Lu 2020 footnote corrected to reflect proper Fig. references and n=4 eyes per condition.

Unverifiable claims:

  • 5fC/5caC discovery by Ito 2011 and TDG excision specificity from He 2011 (both not_oa; chemistry is canonical and uncontested).
  • TET-TDG-BER axis synthesis from Kohli & Zhang 2013 review (PDF download failed; content consistent with verified primary sources).

Downstream propagation needed (main agent):

  • molecules/proteins/tet2.md — if it cites Jaiswal 2014 for TET2-specific prevalence numbers, the framing should be corrected to “overall CHIP prevalence” with TET2 as second-most-frequent driver.
  • phenotypes/clonal-hematopoiesis.md — already verified and uses correct numbers; no change needed.
  • cell-types/hematopoietic-stem-cells.md — already verified and uses correct numbers; no change needed.

[2026-05-07] verify — molecules/proteins/mutyh.md

Pages verified: 1

  • molecules/proteins/mutyh.md — corrections: 7 (see below); 2 sources closed-access

Sources checked:

  • doi:10.1038/ng828 (Al-Tassan 2002, Nat Genet) — local PDF verified (completed download)
  • doi:10.1128/jb.178.13.3885-3892.1996 (Slupska 1996, J Bacteriol) — OA PDF downloaded and verified (PMC232650)
  • doi:10.1128/jb.181.19.6210-6213.1999 (Slupska 1999, J Bacteriol) — OA PDF downloaded and verified
  • doi:10.18632/oncotarget.4284 (Grasso 2015, Oncotarget) — diamond OA PDF downloaded and verified
  • doi:10.1016/j.freeradbiomed.2017.01.008 (Banda 2017, Free Rad Biol Med) — PDF download failed (OA URL resolved to JPEG not PDF); verified via PMC5457711 full text
  • UniProt Q9UIF7 — length (546 aa), active-site residues (D131, D233) confirmed via REST API
  • NCBI Gene 4595 — MUTYH identity confirmed
  • Open Targets ENSG00000132781 — entry exists (445 disease associations, tractability data present); no-opentargets-entry claim was incorrect
  • GenAge — MUTYH not listed (confirmed via direct lookup); genage-id: null correct
  • doi:10.1016/j.dnarep.2006.11.001 (Cheadle 2007) — not_oa; no-fulltext-access tagged
  • doi:10.1172/jci65053 (Sheng 2012) — not_oa; no-fulltext-access tagged

Corrections made:

  1. Mutational signature attribution (Al-Tassan 2002): “particularly in APC (codons 1309, 1450) and KRAS (codon 12)” → corrected to match what Al-Tassan 2002 actually reports: 15/18 somatic APC mutations were G:C→T:A transversions vs 49/503 sporadic (P=2.77×10⁻¹²) and 15/308 FAP (P=7.69×10⁻¹²). Al-Tassan 2002 does not mention KRAS codon 12; KRAS claim removed from Al-Tassan 2002 attribution and flagged unsourced.
  2. Mouse KO evidence (Grasso 2015): “develop intestinal tumors with the G:C→T:A transversion signature” → Grasso 2015 does not demonstrate the transversion signature; it demonstrates adenoma number/severity under AOM/DSS protocol. G:C→T:A signature in Mutyh⁻/⁻ mice comes from other studies. Corrected with actual Grasso 2015 data: n=20/genotype AOM/DSS arm; 35% vs 70% survival (p=0.011 Log-rank); 3/7 KO developed adenocarcinomas vs 0/14 WT.
  3. 8-oxoG lesion frequency: “~10^4–10^5 lesions per cell per day” attributed to Banda 2017 → Banda 2017 reports “once per million guanine residues” (no per-day figure). Attribution removed; figure tagged unsourced pending primary citation (Lindahl 1993 / Fraga 1990 likely sources).
  4. Open Targets claim: “MUTYH has no entry in Open Targets Platform” → incorrect. MUTYH (ENSG00000132781) has an entry with 445 disease associations and tractability labels. no-opentargets-entry removed. Body updated to clarify that “Approved Drug” tractability reflects disease-level associations, not MUTYH-directed drugs; tier 4 retained.
  5. Active site residues: D138/E37 (seeder brief values) — confirmed not found in any source. UniProt canonical (546 aa) gives D131/D233 (confirmed). Banda 2017 uses 535 aa isoform α-3 and reports E120/D222 — consistent 11-aa offset. Page updated to clarify the isoform context of each numbering system.
  6. Grasso 2015 footnote: updated with verified quantitative data (n, survival %, p-value, adenocarcinoma counts, cytokine findings).
  7. Banda 2017 footnote: updated with correct author list, PMC ID (PMC5457711), isoform clarification (535 aa α-3), and note that per-day lesion frequency is not stated in the paper.

Unverifiable claims:

  • Cheadle 2007: ~70–80% MAP alleles Tyr165Cys/Gly382Asp in European populations; ~1–2% CRC prevalence — closed-access; tagged no-fulltext-access
  • Sheng 2012: MUTYH pro-degenerative role, PARP1 hyperactivation in post-mitotic neurons — closed-access; tagged no-fulltext-access

Downstream propagation needed (main agent):

  • pathways/base-excision-repair.md — if it describes the 8-oxoG lesion frequency as 10^4–10^5/cell/day and cites Banda 2017, the attribution should be flagged unsourced (Banda 2017 does not provide this number)
  • hallmarks/genomic-instability.md — if it cites the mouse MUTYH KO for G:C→T:A transversion signature, verify the specific source (Grasso 2015 does not demonstrate this directly)

Final verified state: verified: true with scope (Tahiliani 2009 verified via PMC XML; Globisch 2010 verified against full PDF; Jaiswal 2014 verified against primary PDF; Lu 2020 cross-checked against verified study page; Ito 2011 / He 2011 not_oa; Kohli & Zhang 2013 download failed).

[2026-05-07] verify — microbiome/lactobacillus.md

Pages verified: 1 (partial scope)

  • microbiome/lactobacillus.md — 8 corrections; banner replaced with partial-verification note; verified: true with scope

Sources checked:

  • doi:10.1099/ijsem.0.004107 (Zheng 2020 IJSEM) — abstract via Crossref only (PDF not downloadable, HTTP 403 from publisher); no-fulltext-access for full-text claims
  • doi:10.1038/s41598-023-31115-8 (Liu 2023, Sci Rep) — full PDF verified
  • doi:10.1371/journal.pone.0010667 (Biagi 2010, PLoS ONE) — full PDF verified
  • doi:10.3390/microorganisms9061344 (Hutchinson 2021, Microorganisms) — full PDF verified
  • doi:10.3389/fmicb.2019.02653 — WRONG PAPER (Camardo Leggieri et al., mycotoxin paper); correct DOI for Bui 2019 lactate cross-feeding is 10.3389/fmicb.2019.02449 (confirmed via Crossref); PDF at correct DOI not yet verified
  • doi:10.1016/j.cub.2016.04.016 (Biagi 2016, Curr Biol) — PDF download failed (no candidate URLs); not verified
  • PMID 39505797 (Ji 2025) → doi:10.1007/s11357-024-01419-2 resolved; PDF not verified
  • PMID 38731665 (Chaiyasut 2024) → doi:10.3390/foods13091293 resolved; PDF not verified
  • PMID 38647087 (Zhou 2024) → doi:10.1021/acs.jafc.3c09815 resolved; PDF not verified

Corrections made:

  1. Biagi 2010 cohort size: “n=~90” → n=84 (21 centenarians + 22 elderly + 20 young + 21 centenarian offspring; from Methods section)
  2. Biagi 2010 Lactobacillus claim corrected: wiki said “enrichment of Bacilli-class organisms including Lactobacillus-related taxa” — WRONG. The paper reports Bacilli enrichment driven by Staphylococcus and C. leptum group, with no specific Lactobacillus genus finding. Key centenarian findings are Eubacterium limosum ~15-fold increase, decreased F. prausnitzii, Proteobacteria enrichment. Body and footnote rewritten to match the actual data.
  3. Hutchinson 2021 footnote “819 subjects total” REMOVED — this number does not appear in the paper. Actual trial sizes range 18–1072; no pooled N is reported (no meta-analysis performed). Corrected to “individual trial sample sizes ranged from 18 to 1,072.”
  4. Bui 2019 DOI CORRECTED: 10.3389/fmicb.2019.02653 → 10.3389/fmicb.2019.02449. The original DOI resolves to an unrelated paper on mycotoxin production by Aspergillus/Fusarium.
  5. Liu 2023 MR methods added: primary method GCTA-GSMR (not IVW); sensitivity methods listed. Sample sizes added (N=1539 exposure, N=4477 outcome, N=6548 replication). F-statistic=51.4. β=0.09, p=0.014 for L. amylovorus.
  6. Liu 2023 genus-level Lactobacillus finding added: the genus Lactobacillus (sensu lato) was NEGATIVELY associated with longevity in forward MR; only the species L. amylovorus showed positive association. This nuance was absent from the wiki body.
  7. Ji 2025, Chaiyasut 2024, Zhou 2024 DOIs resolved via PubMed efetch and added to footnotes.
  8. Zheng 2020 footnote updated: note that species count is 261 (not “>250”), PDF unavailable, abstract-only verification.

Unverifiable claims (sources not read end-to-end):

  • Biagi 2016 — Lactobacillus framing in extreme longevity; PDF not downloadable
  • Bui 2019 (correct DOI) — lactate cross-feeding mechanism; PDF at correct DOI not yet verified
  • Ji 2025, Chaiyasut 2024, Zhou 2024 — all PMID-only citations; DOIs resolved but PDFs not read

Downstream propagation needed (main agent):

  • microbiome/bifidobacterium.md — may cite Biagi 2010 for the bifidobacteria decline in centenarians finding (actual paper data: Bifidobacterium significantly lower in centenarians vs young, P=0.023; not absent, just lower). Cross-check footnote.
  • microbiome/gut-microbiome-aging-shifts.md — if it cites Biagi 2010, verify no reference to Lactobacillus enrichment in centenarians (this claim is not supported by the paper).
  • studies/bui-2019-anaerostipes-lactate-crossfeeding.md — if this study page exists, update the DOI to 10.3389/fmicb.2019.02449.

Final verified state: verified: true with scope — see frontmatter verified-scope field.

[2026-05-07] propagation R33 batch 1

Surgical edits propagating R33 batch-1 verifier corrections from 7 freshly-verified pages into upstream/related pages. No re-verification; only edits to remove inherited factual errors. verified-date: on propagated-into pages NOT changed (these are downstream edits).

Pages edited:

  • microbiome/firmicutes-bacteroidetes-ratio.md — Wilmanski 2021: corrected scope of survival analysis to longitudinal MrOS only (N=706 community-dwelling males; 85+ subgroup N=256; MrOS male-only). Both body and footnote updated.
  • molecules/proteins/fgf21.md — Zhang 2012 lifespan-extension paragraph rewritten with sex-stratified medians (males 27.9→36.2 mo ~30%; females 28.8→40.1 mo ~39%; combined 28.1→38.2 mo ~36%); mechanism corrected to JAK2/STAT5 attenuation in GH/IGF-1 axis (NOT mTOR/AMPK/sirtuin); footnote updated.
  • pathways/fgf-signaling.md — Kuro-o 1997 footnote updated: 61-day median lifespan removed from primary statement and tagged no-fulltext-access (closed-access PDF; figure consistent with secondary literature but unverifiable).
  • molecules/proteins/klotho.md — body claim “Premature death at 8-9 weeks” annotated with no-fulltext-access on the widely-cited 61-day median; flagged for re-anchoring to a verified secondary source.

Pages where the correction was already applied (no edits needed; these pages were verified in this batch and already carry the corrected content):

  • microbiome/firmicutes-bacteroidetes-ratio.md — Turnbaugh 2006 mouse-only framing already corrected; Mariat 2009 n=62 details already corrected.
  • microbiome/gut-microbiome-aging-shifts.md — Bian 2017 (n=1,095, age range 3–>94 with primary 883 + secondary 212), Luan 2020 (20-month follow-up / 15-month sampling), Wilmanski 2021 (N=706 MrOS subset), van Soest 2020 (NU-AGE Dutch n=226), Mariat 2009 — all already corrected in verified state.
  • microbiome/bifidobacterium.md — Mariat 2009 (n=62) and Falony 2006 (Anaerostipes caccae DSM 14662 + Roseburia intestinalis DSM 14610) already correct.
  • microbiome/lactobacillus.md — Bui 2019 DOI already corrected to 10.3389/fmicb.2019.02449.
  • pathways/lps-tlr4-nfkb.md — Kim 2016 mouse Firmicutes-INCREASE framing already correct; “study does not report human LPS fold-changes” already noted.
  • processes/gut-barrier.md — Thevaranjan 2017 corrections (18-22 mo mouse age; MDP not LPS as systemic marker; TNF + IL-6; anti-TNF non-rescue of barrier permeability) already applied in verified state.
  • pathways/fgf-signaling.md — Zhang 2012 sex-stratified lifespan, Chang 2024 PKR/mtRNA mechanism, Wedemeyer 2025 closed-access flag, Youm 2025 TEC/adipocyte details — all already correct in verified state.

Greps with no propagation needed:

  • hallmarks/dysbiosis.md and hallmarks/chronic-inflammation.md — do not cite Thevaranjan 2017; LPS framing on dysbiosis page already unsourced (no Thevaranjan-derived claims to correct).
  • hallmarks/altered-intercellular-communication.md — does not cite Zhang 2012, FGF21, or Chang 2024.
  • processes/mitohormesis.md (no Chang 2024 cite; ISR framed via DELE1-HRI generally), hallmarks/mitochondrial-dysfunction.md (no FGF21/Chang 2024 cites). processes/integrated-stress-response.md does not exist.
  • thymus pages — none exist; no propagation needed for Wedemeyer/Youm corrections.
  • molecules/proteins/gdf15.md, molecules/proteins/gcn2.md, interventions/lifestyle/methionine-restriction.md, interventions/lifestyle/ketogenic-diet.md, interventions/gene-therapy/aav-klotho.md, frameworks/interventions-by-modality.md — grep hits were unrelated to Zhang 2012 (the ~30% hit in methionine-restriction.md is the Orentreich 1993 rat MR figure).
  • molecules/compounds/spermidine.md — n=91 hit was Eisenberg 2016 cohort, not Mariat 2009.

[2026-05-07] propagation R33 batch 2

Surgical edits propagating R33 batch-2 verifier corrections from 4 freshly-verified pages (FMT, prebiotics, postbiotics, akkermansia-supplementation) into upstream/related pages. No re-verification; only edits to remove inherited factual errors. verified-date: on propagated-into pages NOT changed.

Pages edited:

  • microbiome/akkermansia-muciniphila.md — Forslund 2015 AKK-metformin paragraph qualified: now reads “to some extent” in the Danish (MHD) cohort only, with cross-cohort inconsistency in CHN and SWE noted; needs-replication appended; verified-scope updated to log the propagation. Other corrections (Everard 2013 dose 2×10^8; Depommier 2019 mean age, HOMA, n=40/32) were already applied in the page’s verified state.
  • hallmarks/dysbiosis.md — postbiotic row in the “Therapeutic strategies” table rewritten to ISAPP-2021-compliant phrasing: pasteurized akkermansia-muciniphila is the leading aging-relevant postbiotic (per Salminen 2021 strict definition); isolated metabolites such as pure SCFAs are NOT postbiotics; urolithin-a explicitly noted as outside the postbiotic class. Replaces prior table row that had conflated postbiotics with “direct metabolite supplementation” (SCFAs as supplements).
  • pathways/lps-tlr4-nfkb.md — “Butyrate / postbiotics” row in the indirect-intervention table rewritten to clarify that butyrate is an isolated SCFA (not a postbiotic per ISAPP 2021); pasteurized A. muciniphila called out as the true aging-relevant postbiotic; Amuc_1100/TLR2 mechanism cited.
  • interventions/dietary/prebiotics.md — Yang 2024 footnote and body updated to reflect: (1) “model: humans ≥65 yr, multicenter (13 communities, Xi’an, China)” replacing “community-dwelling, Xi’an China”; (2) divergent dominant-taxon enrichment by frailty subgroup added (B. adolescentis in prefrail; Oscillibacter sp. in frail).

Pages where the correction was already applied (no edits needed; these pages were verified in this batch and already carry the corrected content):

  • interventions/dietary/fmt.md — Smith 2017 antibiotic-pretreatment-and-recolonization framing (NOT germ-free), 4-8 mo captive lifespan, lifespan numerics (37%/41%/21%) already correct; Bárcena 2019 G609G/G609G genotype, 13.5% / 160 vs 141 days / P=0.0029 / 9% maximal survival / A. muciniphila monoculture already correct; Chen 2020 single 101-yr donor + 70-yr elderly control, 11-mo C57BL/6 mice n=16 (~8/group), expanded SCFA-producer genera (Roseburia, Faecalibacterium, Ruminococcus, Coprococcus) already correct.
  • interventions/dietary/postbiotics.md — Salminen 2021 ISAPP definition, Depommier 2019 mean age ~51 yr / HOMA / no specific LPS percentage, Kang 2024 mean age 65.03±3.83 yr / SARC-CalF<11 / follistatin p=0.0063, Arai 2018 BALB/c mouse strain — all already correct in verified state.
  • molecules/compounds/akkermansia-supplementation.md — Everard 2013 dose 2×10^8, Forslund 2015 cohort qualification, Kang 2024 myostatin null + follistatin p=0.0063 + total extensor significant — all already correct (verifier applied in-place).
  • microbiome/akkermansia-muciniphila.md Everard 2013 dose 2×10^8 already correct (propagated by akkermansia-supplementation verifier).
  • processes/gut-barrier.md — Depommier 2019 already shows HOMA, n=40/32, no stale ~47 yr or clamp wording.
  • model-organisms/nothobranchius-furzeri.md — Smith 2017 details already accurate (antibiotic pretreatment, Ymt/Omt/Abx/wt design, +37%/+41% lifespan).

Greps with no propagation needed:

  • microbiome/akkermansia-muciniphila.md, hallmarks/dysbiosis.md, microbiome/gut-microbiome-aging-shifts.md — no Bárcena 2019 or Chen 2020 references; corrections live only on interventions/dietary/fmt.md.
  • phenotypes/sarcopenia.md, phenotypes/frailty.md — do not cite Kang 2024, PROMOTe, Yang 2024, or Buigues 2016 (those references live only on interventions/dietary/prebiotics.md and interventions/dietary/postbiotics.md); no propagation needed.
  • microbiome/bifidobacterium.md — does not cite Yang 2024 or PROMOTe.
  • pathways/scfa-signaling.md — does not contain “postbiotic” wording; no ISAPP-violation correction needed.
  • studies/lopez-otin-2023-hallmarks-expanding-universe.md, frameworks/hallmark-causality-graph.md — Depommier mentions are qualitative aggregate (no stale numerics); no propagation needed.
  • No file contained the stale wording “germ-free killifish” or “1.5×10^8” outside of log/verified-scope notes.

[2026-05-07] R33 close — Dysbiosis + altered-intercellular-communication

R33 plan: deepen mechanism + intervention coverage of the two integrative hallmarks expanded in López-Otín 2023 (dysbiosis split off as standalone; altered-intercellular-communication carrying microbiome-axis subcomponents pre-2023). Coverage was the thinnest of any hallmark cluster going in.

Pages seeded (11 total, 2 batches, parallel sub-agent dispatch):

Batch 1 — mechanism (7 pages):

  • processes/gut-barrier.md (10 inbound) — tight-junction/mucus/AMP/IgA layered barrier model; aging-permeability shifts; barrier→inflammaging cascade
  • pathways/lps-tlr4-nfkb.md — LBP/CD14/MD2/TLR4 → MyD88 + TRIF arms → IKK → NFκB; aging endotoxin-tolerance paradox
  • pathways/fgf-signaling.md (7 inbound) — endocrine FGF19/21/23 with α/βKlotho cofactors; paracrine FGF1-10; FGFR cascade; FGF21 longevity + Klotho-KO progeria
  • microbiome/gut-microbiome-aging-shifts.md — α-diversity, β-diversity uniqueness (Wilmanski 2021), phylum/genus/species shifts, functional metabolite drift, key cohorts (ELDERMET, NU-AGE, Bian, Wu, Luan)
  • microbiome/bifidobacterium.md — type:microbe genus page (NCBI taxid 1678), bifid shunt biology, lifelong abundance trajectory, prebiotic/probiotic context, strain-vs-species caveat
  • microbiome/lactobacillus.md — type:microbe genus page (NCBI taxid 1578), 2020 Zheng reclassification (25 new genera; many former-Lactobacillus probiotics now elsewhere), Liu 2023 MR genus-vs-species directional split
  • microbiome/firmicutes-bacteroidetes-ratio.md — type:biomarker; honest framing of contested status (Sze & Schloss 2016 meta-analysis null); historical Ley 2005/2006 origin → current consensus to use α-diversity + specific taxa instead

Batch 2 — interventions (4 pages):

  • interventions/dietary/fmt.md — type:intervention, mode:dietary; rCDI-approved (Rebyota Nov 2022 + Vowst Apr 2023) vs aging-investigational; Bárcena 2019 progeroid (13.5% lifespan), Smith 2017 killifish (37%), Chen 2020 centenarian-donor mouse cognition; DeFilipp 2019 ESBL safety alert
  • interventions/dietary/prebiotics.md — class page; ISAPP 2017 canonical definition; PROMOTe 2024 (BCAA+exercise co-intervention; muscle null, PAL cognition p=0.001); Yang 2024 multicenter prefrail/frail subgroup-divergent enrichment; Buigues 2016 nursing-home Fried-criteria improvements
  • interventions/dietary/postbiotics.md — class page; ISAPP 2021 canonical definition (excludes pure metabolites like SCFA); pasteurized Akkermansia Depommier 2019 anchor RCT (HOMA, n=32 completers); Kang 2024 elderly muscle (follistatin p=0.0063, myostatin null); Arai 2018 BALB/c IgA induction
  • molecules/compounds/akkermansia-supplementation.md — type:compound biologic:true; intervention/product layer distinct from microbiome/akkermansia-muciniphila.md organism layer; 14 active ClinicalTrials.gov trials; commercial Pendulum products

Verifier yield (highest of recent rounds — every page had at least one fabrication caught):

  • Zhang 2012 FGF21 lifespan: collapsed sex-stratified figures (males ~30% / females ~39% / combined ~36%) into single bogus number; mechanism wrongly attributed to mTOR/AMPK/sirtuin (actual: JAK2/STAT5 attenuation in GH/IGF-1 axis)
  • Chang 2024 ISR upstream kinase: generic “ISR” → actually PKR responding to cytosolic mtRNA
  • Youm 2025 vs Wedemeyer 2025 thymic FGF21: cellular sources conflated (Youm: TEC + adipocyte; hepatocyte 5x not sufficient)
  • Thevaranjan 2017: 24-mo mouse age (actual 18-22), LPS as systemic marker (actual MDP — LPS only used as ex-vivo macrophage stimulant), anti-TNF rescue inference (paper says no permeability rescue)
  • Kim 2016: fabricated 2-5x human LPS fold-change attributed to a mouse-only paper; Firmicutes direction reversed (actual: Firmicutes INCREASE in aged mice)
  • LPS-TLR4-NFkB: ACCESS phase 3 trial mis-attributed to TAK-242 (actual: eritoran); Bailey 2019 TNFα-only data wrongly extended to IL-6
  • Turnbaugh 2006: misattributed as human-twin study (actual: mouse metagenomics + germ-free transplantation; human twins is Turnbaugh 2009, distinct paper)
  • Mariat 2009: fabricated cohort (n=91 with centenarians) — actual n=62, no centenarians
  • Falony 2006: three fabricated cross-feeding species (Roseburia inulinivorans, Faecalibacterium prausnitzii, Eubacterium hallii) — actual two were Anaerostipes caccae DSM 14662 + Roseburia intestinalis DSM 14610
  • Bui 2019: wrong DOI (10.3389/fmicb.2019.02653) resolved to a completely different mycotoxin paper; correct DOI 10.3389/fmicb.2019.02449
  • Hopkins 2001: “large-scale” framing → actual tiny groups (n=10/7/5/4)
  • Biagi 2010 Lactobacillus enrichment in centenarians: entirely fabricated (paper does not report this)
  • Hutchinson 2021: fabricated “819 subjects total” (number doesn’t appear in paper)
  • van Soest 2020 NU-AGE: “>1,200 pan-European” → actual 226 Dutch-only
  • Bian 2017 age range: “>100” → actual ≥94 (paper never claims subjects over 100)
  • Luan 2020: 15-mo follow-up → actual 20-mo follow-up (15-mo sampling window)
  • Wilmanski 2021 survival scope: implicitly cross-cohort → actual MrOS-only, exclusively male
  • Smith 2017 killifish: “germ-free” → actual antibiotic-pretreated then recolonized
  • Bárcena 2019: missing 13.5% / 160 vs 141 days lifespan numerics; genotype notation (G609G/G609G homozygous)
  • Chen 2020: “centenarian donors” plural → actual single 101-yr donor + single 70-yr elderly control
  • Depommier 2019: ~47 yr → actual ~51 yr; “hyperinsulinaemic clamp” → actual HOMA; fabricated ~35% LPS reduction (paper does not state a specific %)
  • Everard 2013: dose 1.5×10^8 → actual 2×10^8 cfu/day
  • Kang 2024: fabricated myostatin claim — actual myostatin not significantly changed (P=0.3145), only follistatin significant
  • Forslund 2015 AKK-metformin: overstated single-cohort association as cross-cohort; actual finding was Danish-only with CHN/SWE inconsistent
  • Guan 2023: DOI 10.1039/d3mo00068a fabricated (does not exist); correct DOI 10.1039/d2mo00256f
  • Arai 2018 mouse strain: C57BL/6 → actual BALB/c (Th2-biased; immunologically distinct)

Schema escalations:

  • interventions/dietary/ directory was created during R33 batch-2 seeding; CLAUDE.md directory map currently lists only lifestyle/, pharmacological/, gene-therapy/, stem-cell-therapy/, blood-product/ under interventions/. The map should be updated to include dietary/. (R33b deferred; non-blocking.)
  • type: process schema lacks a literature-checked-through: field. The newly-seeded gut-microbiome-aging-shifts.md covers a fast-moving literature class similar to type: compound and type: intervention (which carry the field). Recommend adding literature-checked-through: (optional) to the type:process schema in CLAUDE.md so the lint pass can flag stale process pages on the same cadence as compounds/interventions. (R33b deferred; non-blocking.)

Workflow note: propagation passes were unusually light — both batches’ verifiers had Read+Edit access and applied corrections to the seeded pages in-place during their own runs, leaving the propagation step to handle only the inherited downstream cases (4 hits in batch-1 propagation, 4 in batch-2 propagation). This is the cleanest verifier→propagation handoff we’ve seen; the Plan-mode seeder/verifier brief discipline is paying off.

Remaining R33 items (deferred):

  • None for the dysbiosis/AIC content cluster proper.
  • Schema cleanup (CLAUDE.md directory map + type:process literature-checked-through field) deferred to R34 or a dedicated schema-cleanup batch.
  • Open propagation backlog from R33 verifiers: faecalibacterium-prausnitzii implicit stub (multiple R33 pages reference it as a key butyrate-producer with aging-decline; should be seeded as a microbe page in a future round).

Forward-queue surfacing from R33:

  • TLR4 / TLR2 / MD2 / CD14 / LBP / MyD88 / TRIF / IRAK1 / IRF3 — implicit stubs from lps-tlr4-nfkb.md (most have existing pages but should be cross-checked for adaptor-arm coverage)
  • Faecalibacterium prausnitzii — major butyrate-producing aging-decline taxon; multiple inbound references from R33 pages
  • Roseburia / Eubacterium / Coprococcus genera — referenced by FMT outcome data
  • Amuc_1100 protein — Akkermansia outer-membrane protein, TLR2 ligand; mechanism centerpiece of postbiotic claims
  • Klotho/FGFR1c/FGFR4 individual receptor pages — flagged by fgf-signaling.md as implicit stubs
  • Tricellulin (a tight-junction protein) — flagged by gut-barrier.md
  • Reg3γ / MUC2 / claudin family individual pages — flagged by gut-barrier.md

Graph View

Backlinks