R36

log/R36.md — Round 36 entries

Sub-file of log — see parent for index.

[2026-05-19] verify | cell-types/langerhans-cells.md

Pages verified: 1

• cell-types/langerhans-cells.md — R38 substrate batch. Load-bearing primary sources verified: Merad 2002 (PMC4727838 full text), Bhushan 2002 (PubMed abstract PMID 11841364; not_oa), Pilkington 2018 (PMC5721243 full text), Wang 2009 (PMC2921127 full text + PubMed abstract), Xu 2012 (publisher full text), Cumberbatch 2002 (PubMed abstract PMID 11985667; not_oa), Hochberg 1999 (PubMed abstract PMID 10568168; DOI resolved via Crossref), Mielcarek 2014 (PMC4149838 full text). Reviews (Chorro 2009, Hoeffel 2012, Grewe 2001, Chambers 2020, Pilkington 2021, Toews 1980) not independently re-read; claims from these are contextual/non-quantitative.

Corrections applied (8):

1. “10–15 year half-life in humans” claim removed — Merad 2002 makes no such statement; replaced with actual Merad 2002 findings: >97% host LCs at 18 months, BrdU 5–10% over 10–12 days; gap tag added for human lifespan quantification.
2. Density figure intro: “~30–50%” → “~28%” (matches Bhushan 2002 exact data); skin site added (buttock, not general).
3. Density section body: “sun-protected forearm skin” → “sun-protected buttock skin” (Bhushan 2002 used buttock); exact SEM and p-value added; elderly mean age corrected from “~73y” to “76y (range 72–79)”.
4. Migration section: “TNF-α-induced trauma” → “intradermal TNF-α injection (200 U)” — Bhushan 2002 used pharmacological TNF-α injection, not physical trauma; “trauma” is the Pilkington 2018 model.
5. Wang 2009 mouse model: “Langerin-DTA mice” → “Lang-DTR mice” (DTR = diphtheria toxin receptor, not DTA; confirmed in PMC2921127); UV suppression effect quantified (~5-fold OT-I reduction) added.
6. Hochberg 1999: DOI resolved to 10.1111/j.1751-1097.1999.tb08281.x (from Crossref; previously PMID-only); correct authors (Hochberg M, Enk CD — not “et al.”); “50–70% LC depletion” claim detached from Hochberg 1999 (paper addresses sunscreen protection, not depletion magnitude); gap tag added.
7. Bhushan 2002 footnote: elderly mean age corrected ~73y → 76y; skin site corrected to buttock; exact SEM values added; CD1a staining method corrected (originally said HLA-DR+, actually CD1a immunofluorescence).
8. [[inflammaging]] broken wikilink resolved to inline text + [[chronic-inflammation]] (inflammaging.md does not exist).

Additionally:

• Xu 2012 frontmatter single-cell-aging-signature updated with verified quantitative LC frequency decline figures and miRNA identities.
• Pilkington 2018 footnote: n values added (young n=39, aged n=35); migration percentages added (~15% vs ~4%).
• Wang 2009 footnote: model name corrected; n added; quantitative effect added.
• Mielcarek 2014 footnote: chimerism kinetics added (day 28: 7%, day 84: full donor).

Supersession check: SKIPPED — type: cell-type exempted per CLAUDE.md R25 rules.

Downstream propagation needed:

• hallmarks/disabled-adaptive-immunity.md — may cite LC density/migration claims; check for “forearm skin” site attribution or “10–15 year half-life” language.
• tissues/skin.md — Chambers 2020 note at log entry 2026-05-19 previously flagged LC density as #gap/needs-replication; that gap note may now be partially resolved by Bhushan 2002 data confirmed here.
• Any page citing Bhushan 2002 density should reflect buttock site and 835.7 ± 48.2 cells/mm² elderly value.

[2026-05-19] verify | tissues/skin.md

Pages verified: 1

• tissues/skin.md — R38 substrate batch. PDFs read: Ganceviciene 2012 (all 12 pages), Krutmann 2017 (all 10 pages), Bormann 2016 (8 pages), Paus 2024 (3 pages). Horvath 2013 cross-checked against verified study page. Chambers 2020 and Ho 2021 PDF downloads failed; Pilkington 2021 not_oa. All 5 R38 sister cell-type pages and all phenotype wikilinks confirmed resolvable.

Corrections applied (5):

1. Body-weight/surface-area claim: [^ganceviciene2012] citation removed; #gap/unsourced added — Ganceviciene 2012 is a therapeutics review containing no such anatomy figures.
2. Dermis thickness loss rate (~20%/decade): [^ganceviciene2012] removed from anatomy table; #gap/unsourced added; Shuster 1975 / Gniadecka 1994 flagged as candidate correct sources.
3. Interfollicular melanocyte density (~6–8%/decade): [^ganceviciene2012] removed; #gap/unsourced added; Gilchrest/Fitzpatrick literature flagged as likely source.
4. Bormann 2016 footnote: page numbers corrected 506–516 → 563–571; study design corrected “in-vitro/observational” → “observational”; MAE figures added.
5. Ganceviciene 2012 footnote: description corrected to therapeutics-review framing; explicit note added that anatomy figures were incorrectly attributed.

Added gap tags: #gap/needs-replication on Langerhans cell 20–40% density figure (Chambers 2020 full PDF unavailable).

Downstream propagation needed: check other R38 cell-type pages for Ganceviciene 2012 citations used as anatomy references.

[2026-05-19] verify | cell-types/melanocyte-stem-cells.md

Pages verified: 1

• cell-types/melanocyte-stem-cells.md — R38 substrate batch. Four sources verified against local PDFs: Zhang 2020, Sun 2023, Nishimura 2002, Inomata 2009 (downloaded during session). Nishimura 2005 not_oa; Schouwey 2007 download failed.

Corrections applied (7): see verified-scope in page frontmatter.

Downstream propagation needed:

• phenotypes/hair-greying.md — may cite Inomata 2009 for ATM mechanism; check for “suppressed in Atm-null” framing and correct direction if present.
• Any seeded Inomata 2009 study page should reflect ATM-sensitization (not suppression) direction.

[2026-05-19] verify | cell-types/melanocytes.md

Pages verified: 1

• cell-types/melanocytes.md — R38 substrate batch. Seven primary sources checked: Hattori 2004 (local PDF), Lin & Fisher 2007 (local PDF), Yu 2025 (local PDF), Park 2023 (downloaded gold OA), Imokawa 2019 (downloaded gold OA), Hari 2012 (downloaded OA), Paus 2024 (downloaded bronze OA). Ortonne 1990 and Kim 2025 remain not_oa (#gap/no-fulltext-access retained). Abdel-Malek 2008 download failed (#gap/no-fulltext-access added to footnote).

Corrections applied (8):

1. Hattori 2004 footnote n: “n=12 patients” → “n=29 patients total (RT-PCR n=7, IHC n=10, western blot n=6)”; clarified exact p-values and fold-changes per assay; corrected journal page range to 1256–1265; added c-KIT mRNA unchanged (1.16-fold, NS).
2. Solar lentigines table: “n=12 patients” → “n=29 patients; SCF RT-PCR n=7, IHC n=10, western blot n=6”.
3. Park 2023 body text: “KIF5B/KLC2 expression” → “MYO5A/RAB27A/MLPH tripartite complex” (paper’s actual finding; KIF5B/KLC2 not named as key effectors in Park 2023).
4. Park 2023 footnote: corrected transport-protein names, added UV dose/model detail (60 mJ/cm² × 4 days), updated DOI lookup from “download pending” to confirmed local path; added in-vivo IGH corroboration.
5. Lin & Fisher 2007 “30–40 keratinocytes” claim: ratio does not appear in the paper; added unsourced and attribution note (Fitzpatrick 1971 origin); softened body text and Identity table accordingly.
6. Yu 2025 frontmatter single-cell-aging-signature: corrected from “transfer impaired” to accurate description of eumelanin synthesis upregulation (paper’s actual finding); corrected “snRNA-seq” to integrated scRNA-seq + spatial transcriptomics.
7. Yu 2025 body text: updated to reflect correct finding (elevated eumelanin synthesis, haphazard pigmentation) vs the seeder’s misframing (transfer impairment); added cell count (116,627 cells).
8. Hari 2012: footnote updated with correct journal volume/pages (Development 139(12):2107–2117), DOI lookup added, temporal-control nuance added to body text.

Unverifiable claims: Ortonne 1990 (8–20%/decade density decline) and Kim 2025 (BCL-xL senolytic) — not_oa; no-fulltext-access retained. Abdel-Malek 2008 download failed — no-fulltext-access added. MC1R/CPD claims attributed to Abdel-Malek 2008 unverified against source.

Supersession check (R25): No superseding meta-analysis or large RCT found for solar lentigo SCF/KIT mechanism, melanocyte senescence, or melanocyte-senolytic claims (PubMed search, 2023–2026). literature-checked-through: 2026-05-19 set.

Downstream propagation needed: tissues/skin.md uses “6–8%/decade” for melanocyte density decline while melanocytes.md retains Ortonne “8–20%/decade” — inconsistency between pages; main agent should reconcile when Ortonne 1990 or a modern replacement source becomes available.

[2026-05-19] verify | cell-types/dermal-fibroblasts.md

Pages verified: 1

• cell-types/dermal-fibroblasts.md — R38 substrate batch. Five primary sources verified against PDFs: Demaria 2014 (existing local copy), Velarde 2012 (existing local copy), Fisher 1996 (existing local copy), Fisher 2009 (downloaded from bronze OA URL), Solé-Boldo 2020 (downloaded from PMC gold OA). Purohit 2016 unverifiable (not_oa); no-fulltext-access retained.

Corrections applied (5):

1. Sole-Boldo 2020 subset names: “Papillary-like / Reticular-like / MPS-like” → “Secretory-papillary / Secretory-reticular / Mesenchymal” (paper uses these exact cluster names; Table 1 + Figure 4a legend)
2. Sole-Boldo frontmatter single-cell-aging-signature field: updated to correct subset names
3. Sole-Boldo footnote: added n (5 donors, 2 young/3 old), donor ages, cell counts, spatial-validation note; removed “download pending” status
4. Fisher 2009 YAP/TAZ misattribution: Fisher 2009 does NOT describe the YAP/TAZ pathway — the mechanosensing mechanism it describes is c-Jun/AP-1 + α2β1 integrin. Corrected mechanosensing section (body) and niche-signalling YAP/TAZ section. Added unsourced for the YAP/TAZ claim in aged DFs in vivo. Added Note to fisher2009 footnote: “YAP/TAZ pathway is NOT described in this paper.”
5. Fisher 2009 loop description: removed “YAP/TAZ cytoplasmic” from step 4 of the collagen fragmentation loop; replaced with the paper’s actual mechanism (c-Jun/AP-1 + α2β1 integrin + ROS). Added quantitative detail (MMP-1 mRNA ~8-fold elevated in vivo; ROS ~3-fold in fragmented lattice; MitoQ10 reduces MMP-1 ~30–40%).

Unverifiable claims: Purohit 2016 (doi:10.1016/j.jdermsci.2016.04.004) — Smad3/TGF-β quantitative claims unverifiable; not_oa. no-fulltext-access retained.

Supersession check: No newer meta-analysis or large RCT found for senolytic skin aging endpoints (as of 2026-05-19). Newer scRNA-seq papers (2021–2026) do not supersede Sole-Boldo 2020 for the aging-dermal-fibroblast-subset question. literature-checked-through not applicable to type:cell-type.

Downstream propagation needed: phenotypes/skin-aging.md — references Fisher 2009 for the collagen fragmentation loop; if it attributes the mechanosensing step to YAP/TAZ via Fisher 2009 it should also be corrected (verified page; defer to main agent).

[2026-05-19] verify | tissues/dermis.md

Pages verified: 1

• tissues/dermis.md — R38 substrate batch. All six primary sources verified against local PDFs: Fisher 2009 (downloaded from OA URL, bronze OA), Fisher 1996 (existing local copy), Demaria 2014 (existing local copy), Velarde 2012 (existing local copy), Fisher/Voorhees 2023 (downloaded, green OA via PMC10409944), Shin 2019 (downloaded, gold OA via PMC6540032).

Corrections applied (6):

1. Collagen I dry-weight proportion: “~70% dry weight” → “~60–68% dry weight” (collagen fibers = ~75% dry weight; collagen I = 80–90% of that; per Shin 2019)
2. Collagen III dry-weight proportion: “~10–15% dry weight” → “~6–9% dry weight” (8–12% of total collagen × ~75%; per Shin 2019)
3. HA intrinsic-aging claim: “reduced HA content with age” → corrected — per Shin 2019, HA quantity is NOT significantly different in intrinsic aging; HA-binding proteins (HABPs) decline. Photoaged skin shows increased HA in solar elastosis regions. Gap tag adjusted accordingly.
4. Versican intrinsic-aging: “Upregulated in aged dermis” → “status uncertain” — Shin 2019 Table 1 records “Not changed?” for intrinsic aging; tagged contradictory-evidence
5. “~1%/yr collagen loss” figure: removed implicit attribution to verified sources; tagged unsourced across all three locations where it appeared (ECM table, aging changes table, intrinsic vs photoaged table)
6. Reticular dermis MMP feedback loop: clarified that Fisher 2009’s demonstrated loop is MMP-1-centred (not MMP-1/3/9 jointly); MMP-3 and MMP-9 are secondary processors of cleavage products, not co-drivers of the feedback cycle per Fisher 2009’s experimental results.

Unverifiable claims: Citation counts (444 citations for Fisher 2009; 1,395 for Fisher 1996; 1,889 for Demaria 2014; 724 for Shin 2019; 63 for Fisher 2023) are a local paper DOI lookup, not verifiable from PDFs. LOX activity decline primary source (Cenizo et al. — cited in Shin 2019 review) not directly verified.

Downstream propagation candidates:

• phenotypes/skin-aging.md — contains the “~1%/yr” collagen figure in its Dermal Changes section (line: “estimated ~1% per year of adulthood”) citing [^fisher2009]; this figure is unsourced and should be tagged there too. Also check HA framing.
• tissues/epidermis.md — may cite Fisher 2009 or Shin 2019 for collagen figures; check for consistency

[2026-05-19] verify | cell-types/keratinocytes.md

Pages verified: 1

• cell-types/keratinocytes.md — R38 substrate batch.

Sources checked: Fisher 1996 (10.1038/379335a0), Hattori 2004 (10.1111/j.0022-202x.2004.22503.x), Velarde 2012 (10.18632/aging.100423), Clayton 2007 (10.1038/nature05574), Mascré 2012 (10.1038/nature11393) — all verified against local PDFs. Rheinwald 1975 (not_oa) and Kovacs 2010 (not_oa) unverifiable.

Corrections applied (6):

1. tissue-of-origin: [[skin]] → [[epidermis]] (race-condition fix; epidermis.md now exists)
2. Clayton 2007 symmetric division probability: “~10%” → “~8%” (paper: r = 0.08; rλ = 0.088 ± 0.004/week)
3. Mascré 2012 body: “K14^low / K14^high” nomenclature replaced with correct K14-Cre-ER / Inv-Cre-ER population description; wound-healing contribution detail added from paper
4. Mascré 2012 footnote: “K14-rtTA + K10-CreERT” → “K14-Cre-ER and Inv-Cre-ER” (paper uses involucrin-Cre-ER, not K10-CreERT)
5. Velarde 2012 footnote: n values corrected and stratified (aged C57BL/6J mice n=6/10/13 by age group; Sod2-/- Western n=6/6; histology n=8/9); cell lines named explicitly (AG21837 keratinocytes, HCA2 fibroblasts)
6. Fisher 1996 tretinoin claim: “~70%” AP-1 reduction and “50–80%” MMP reduction removed as unsupported explicit percentages; replaced with reported significance levels (p<0.01) and gap tag
7. Hattori 2004 footnote: “29 Japanese patients” removed (not stated in paper; patient count varies per assay; figures number patients to at least 27); GROα n corrected to n=4 (RT-PCR), bFGF n=10 (IHC)

Downstream propagation candidates:

• phenotypes/skin-aging.md — may have inherited Mascré 2012 K14^low/K14^high language or Fisher 1996 “~70%/50–80%” figures
• studies/fisher-1996-photoaging-ap1-mmp (if seeded) — verify AP-1 reduction percentages
• studies/hattori-2004-scf-solar-lentigo (if seeded) — verify patient count

[2026-05-19] verify | tissues/epidermis.md

Pages verified: 1

• tissues/epidermis.md — R38 substrate batch. Fisher 1996, Velarde 2012, Hattori 2004 verified against primary source PDFs. Grove 1983, Kligman 1993, Ortonne 1990, Wulf 2004 unverifiable (all not_oa); existing no-fulltext-access tags confirmed correct.

Corrections applied (1):

• Transit-time body text: “30–35 days in older adults” → “>30 days in older adults” — the “30–35” upper bound was not supported by the Grove 1983 footnote, which documents only “>10 days longer” (i.e., “>30 days”); closed-access paper prevents exact figure verification.

No corrections needed for PDF-verified sources:

• Fisher 1996: n ranges (6–17 mRNA; n=9 AP-1/NF-κB; n=10 protein/activity), MMP-1/3/9 induction, MMP-2 non-induction, AP-1 binding ~70% reduction, MMP activity 50–80% reduction, transrepression mechanism — all confirmed.
• Velarde 2012: p16 ~2-fold elevation, epidermal thinning in Sod2-/- (CD1 background, EUK-189), SA-bgal in stratum corneum, keratinocyte rotenone sensitivity > fibroblast — all confirmed.
• Hattori 2004: SCF mRNA 3.9-fold n=7 p<0.01; SCF protein 1.6-fold n=6 p<0.05; GROα 1.13-fold NS n=4; bFGF NS IHC n=10; 29 Japanese patients ages 35–85 — all confirmed.

Wikilinks: all 12 checked — all resolve to existing pages.

Downstream propagation needed: none (no numerical corrections to PDF-verified sources; the transit-time correction is minor and the claim is already tagged needs-replication + no-fulltext-access on the epidermis page. skin-aging.md already uses “>30 days” phrasing in its own body — no conflict).

[2026-05-18] ingest | berberine compound page + 7 study pages + 5 propagation passes

Trigger. User-supplied: empty placeholder berberine.md at the repo root (zero bytes), 9 study citations (Zhu 2015, Ma 2022, Lei 2026 BRAVO, Kong 2004, Zheng 2025, Derosa 2013, Yang 2023, Chen 2020 + Tan 2025 follow-up), and two YouTube transcripts — one on plaque reversal and one on colorectal-adenoma prevention. Ad-hoc seed (per feedback_adhoc_seed_no_roadmap memory); not added to ROADMAP.

Pages created (8).

• molecules/compounds/berberine.md — type: compound, 365+ lines, ~15 footnotes; mechanism / primary-effect / downstream-effect strictly separated per user request. Frontmatter verified: true with partial scope (Ma 2022 PDF + BRAVO/Tan 2025 PMC primary-verified; Kong/Yang/Derosa/Zhu/etc. secondary-cited pending full PDF). Removed empty placeholder at repo root.
• studies/ma-2022-berberine-tmao-atherosclerosis.md — verified: true end-to-end from local PDF.
• studies/lei-2026-bravo-berberine-masld.md — verified: true via PMC12811813 full text (PDF still pending download).
• studies/tan-2025-berberine-adenoma-6yr-followup.md — verified: true via PMC12490254 (PDF pending).
• studies/chen-2020-berberine-adenoma-rct.md — abstract-verified via PubMed efetch; verified: false + #gap/no-fulltext-access (Lancet GH not_oa).
• studies/kong-2004-berberine-ldlr-mechanism.md — verified: false; not_oa + no PMC version.
• studies/yang-2023-berberine-cv-meta.md — verified: false; not_oa + no PMC version.
• studies/zheng-2025-berberine-vsmc-jak2-stat3.md — verified: false; pending download.

Corrections from verification (important).

• Ma 2022 model is golden hamster, not ApoE−/− (widely misattributed in secondary literature).
• Ma 2022’s significant human endpoint is composite carotid plaque score (n=21, −3.2%, p<0.05 one-tailed), NOT carotid IMT (3.2% change but p=0.067, NS). Secondary sources including the user’s YouTube transcript conflated the two.
• Berberine does NOT directly inhibit CutC — gut-bacterial nitroreductase (NR) reduces berberine to dihydroberberine (dhBBR), and dhBBR is the active inhibitor of bacterial CutC + a gut-bacterial FMO. Hepatic FMO3 itself is NOT inhibited by berberine (Ma 2022 confirmed in liver homogenate assays). Substantive refinement of the brief-stage framing.
• Chen 2020 CBAR dose: 300 mg BID = 600 mg/day for 2 years (confirmed via PubMed efetch + Tan 2025 PMC). The Tan 2025 seeder’s initial WebFetch extraction transcribed the dose as TID; corrected on both pages.
• Chen 2020 evaluable n=891 (429 berberine + 462 placebo), not “~1000” as widely reported. 895 enrolled.
• Tan 2025 6-yr follow-up: any-adenoma adjusted HR 0.58 (95% CI 0.45–0.74, p<0.001), effect strengthens over time; advanced adenoma HR 0.99 (p=0.57) — NULL (only 13 vs 17 events; underpowered). The advanced-adenoma null is an important caveat the second YouTube transcript did not emphasize.
• BRAVO 2026: n=337 randomized, co-primary endpoints negative (VAT diff +1.38% p=0.42; liver fat diff +0.87 pp p=0.12). Secondary endpoints positive: LDL-C −7.72 mg/dL (p=0.008), ApoB −3.42 mg/dL (p=0.02), hs-CRP −0.072 mg/dL (p=0.04). Both arms received lifestyle counseling and both lost ~1.9 kg — the active-comparator design eliminates between-arm body-composition separation but preserves the mechanism-driven lipid signal.

Propagation (6 pages updated, plus intervention-classes schema additions).

• interventions/pharmacological/ampk-activators.md — compound-table row + safety paragraph + aging-data line refreshed with Chen 2020 / Tan 2025 / BRAVO numerics + the “AMPK-activator classification is mechanistically incomplete for berberine’s strongest human evidence” caveat.
• microbiome/akkermansia-muciniphila.md — berberine bullet upgraded from #gap/unsourced to Ma 2022 cited with mechanism detail; added Ma 2022 footnote.
• molecules/compounds/akkermansia-supplementation.md — berberine bullet split out of the generic-polyphenols line; documents NCT07440147 AKK+berberine combination trial.
• phenotypes/atherosclerosis.md — new “Gut-microbiome-axis modulation — berberine” subsection under § Pharmacological interventions; converges three mechanisms (LDLR mRNA, TMAO, JAK2/STAT3) with honest framing of human-IMT evidence limitations.
• microbiome/gut-microbiome-aging-shifts.md — § TMAO expanded with the prototype-berberine “vitamin-like luminal agent” framing; added Ma 2022 footnote.
• hallmarks/dysbiosis.md — § Therapeutic angles table now includes berberine row.

Schema escalations (handled by compound seeder). Added four new mechanism sub-class aliases to frameworks/intervention-classes.md: ldlr-mrna-stabilization, pcsk9-reduction, tmao-suppression-microbial, jak2-stat3-inhibition, plus a new top-level class glp1-secretion-stimulation (distinct from GLP-1R agonist class). All annotated with “Added 2026-05-18 (first compound page: berberine)”.

Gaps surfaced.

• #gap/needs-replication — Most IMT/plaque RCT data lives in Chinese-language regional journals; Yang 2023 meta-analysis’s 13 IMT-subgroup component RCTs largely not PubMed-indexed (independently corroborated via PubMed search). Non-Chinese-population atherosclerosis-outcome RCT for berberine is absent.
• #gap/no-mechanism — Durability of CRC chemoprevention effect 6 years after berberine discontinuation (Tan 2025) is uncharacterized mechanistically; microbiome reprogramming is the leading hypothesis but unconfirmed.
• #gap/no-mechanism — Bioavailability paradox: how does 0.68% oral absorption produce systemic LDL-C effects? Trace hepatic exposure, active metabolites (berberrubine), and enteric AMPK all candidate explanations.
• #gap/contradictory-evidence — BRAVO adiposity null vs prior smaller-trial meta-analytic adiposity-reduction signals; lifestyle-comparator confound is the most parsimonious explanation but not formally tested.
• #gap/long-term-unknown — Beyond 6 years; no ITP lifespan data; no analog of TAME for berberine.
• clinical-trials-active: 40 on the compound page is an order-of-magnitude estimate; exact ClinicalTrials.gov v2 API query not performed on this date. #gap/needs-ct-gov-count-refresh

---

[2026-05-18] verify | tretowicz-2026-blood-nad-stable-aging + nad-blood-biomarker (full-PDF cross-check)

User supplied the 24-page Trętowicz 2026 Nat Metab PDF (DOI 10.1038/s42255-026-01537-5) via local Downloads — the same paper that was seeded on 2026-05-15 from Abstract + Reporting Summary + 8 source-data XLSX files only (paywalled Results/Discussion). Full PDF now read end-to-end; the 2026-05-15 extraction is confirmed accurate on every quantitative claim cross-checkable to the now-visible Results, Methods, and Extended Data sections.

Pages verified (verified: false → true):

• studies/tretowicz-2026-blood-nad-stable-aging.md — verified-by: claude; verified-date: 2026-05-18; banner removed
• biomarkers/nad-blood-biomarker.md — verified-by: claude; verified-date: 2026-05-18; banner removed

Corrections applied during verification:

• Study page cohort table: ELITE trial-registry row was duplicating the LLS identifier (NL-OMON52307) — corrected to NL71682.018.19 / NL9328. MEJNES2019 NCT (NCT06975540) added; METc ID typo corrected to H150037557317.
• TEAMS subgroup n: paper Results-text says n=34 exercise+protein; Extended Data Table 1 says n=35. Cohort table now cites Results-text value with explicit note of the minor discrepancy.
• Method-validation parameters added to both pages (now PDF-visible): LOD 0.21 nmol/mL, LOQ 0.69 nmol/mL, intra-assay CV 12.5%, inter-assay CV 16.1%, linearity R²=0.9996, spike-recovery 108–112%. Establishes the assay’s analytical noise floor — the null age effect is well above the floor.
• Storage-temperature correction: prior pages emphasized only −20°C as degrading; full Results text confirms −80°C also significantly degrades NAD+ (fresh ~40 → −80°C ~31 nmol/mL, P=0.0015; vs −20°C ~24 nmol/mL, P=0.023). Both pages updated to reflect that no standard freezer protocol preserves NAD+ fidelity.
• NR positive-control quantitative values added: 32 → 62 nmol/mL median pre→post, P=1.59×10⁻⁹ (Extended Data Fig. 3) — previously cited only qualitatively as “increased as expected.” Twin-pair cohort population details added: 24 BMI-discordant monozygotic twins, ages 33–41 y, FinnTwin12/16 parent cohort, NR escalated 250 mg/day → 1 g/day over 5 months.
• Freeze-thaw per-donor variability nuance: full PDF reveals Fig. 1f shows two representative donors with cumulative cycle-3 losses of −4% (with mid-cycle rebound) vs −32%; previous framing of “30–80% loss” is correct as a worst-case bound but obscures the within-donor recovery and inter-donor variance. Mechanistic implication added — freeze-thaw NAD+ loss is not deterministic per cycle.
• Study page methods § Plasma updated: explicit “50–100× lower than whole-blood” + “below LOQ (0.69)” — previously only cited the absolute nmol/mL values.

Downstream propagation status (already done on 2026-05-15, NOT re-touched in this pass):

• interventions/pharmacological/nad-precursors.md — R37 entry already integrated Trętowicz 2026 into “NAD+ Declines with Age” section, the extrapolation table, the class-level 2024–2026 evidence-wave summary, and footnotes; literature-checked-through 2026-05-15.
• molecules/compounds/nr.md — already integrated into “restore-deficit rationale undermined at blood level” caveat + footnote.
• molecules/compounds/nmn.md — already integrated into the age-decline-tissue-specific paragraph + footnote.

No new propagation needed — the 2026-05-15 seeder pass already touched every downstream page that should reference this paper.

Critique of the paper (recorded for future readers): Methodologically rigorous — 7-cohort multi-population replication, embedded positive control, pre-analytical artifacts characterized, power analysis included, no competing interests. Three reviewers (Covarrubias, Migaud, Wu) include known NAD+-field methodological skeptics. The result is the most rigorous human test of blood-NAD+/aging to date. Key caveat the paper acknowledges: whole-blood NAD+ ≠ tissue NAD+ (RBCs are anucleate, 120-day turnover); the negative result is specific to blood as a biomarker, NOT to tissue NAD+ decline as a biological phenomenon. Janssens 2022 (human muscle) and Massudi 2012 (human skin) tissue-level decline findings remain valid in light of this paper.

Reconciliation of prior contradictions: The paper directly explains the discrepancy with prior positive findings (Chaleckis 2016 PNAS, Wang 2023, Yang 2022, Breton 2020) as pre-analytical handling artifacts — frozen-without-methanol-preservation samples lose 30–80% of NAD+ depending on conditions. Differential handling across age groups (e.g., older participants more often sampled in clinical settings with different storage logistics than younger volunteers) could spuriously generate age-correlated NAD+ “decline” in poorly-controlled studies. The wiki’s framing of this contradiction in biomarkers/nad-blood-biomarker.md § “Sources of confusion in the prior literature” is consistent with the paper’s own explanation.

[2026-05-15] verify+propagate | Yang 2003 footnote + GlycoSENS cross-references

Follow-on pass to the same-day AGE-breaker landscape update (see entry below). Verified the load-bearing Yang 2003 critique and propagated downstream.

Yang 2003 verification (abstract-level via PubMed):

• PMID:12646266 · Yang S, Litchfield JE, Baynes JW · Arch Biochem Biophys 2003 Apr 1;412(1):42–6
• a local paper DOI lookup confirmed: oa_status closed, download_status not_oa, citation_percentile 99.0, fwci 4.64 (high-impact, closed-access — full PDF retrieval not currently possible via archive)
• Full abstract retrieved via PubMed WebFetch; verification status flipped from “abstract-level — verification pending” to “VERIFIED at abstract-level via PubMed 2026-05-15”
• New facts surfaced from the full abstract (not in prior wiki framing):
  • Compounds tested: PTB, PMT (N-phenacyl-4,5-dimethylthiazolium — essentially the ALT-711/alagebrium pharmacophore), AND pyridoxamine (PM). Prior wiki framing said only “PTB and related breakers”; PMT/alagebrium-chemistry inclusion is a stronger critique than initially captured.
  • Pyridoxamine also failed to cleave established crosslinks in the same study. The wiki had not previously distinguished pyridoxamine’s upstream carbonyl-trap mechanism from a crosslink-reversal capability; updated advanced-glycation-end-products pyridoxamine subsection 2026-05-15 to make this explicit.
  • Quantitative anchors: diabetic skin collagen ~5× nondiabetic crosslinking burden; diabetic tail tendon ~10×; methods: PBS pH 7.4 × 24h × 37°C; readouts: pepsin solubility (skin) + acid solubility (tail).
  • Model crosslink used: phenylpropanedione.
  • Direct verbatim conclusion quote captured in footnote.

Propagation downstream (completed 2026-05-15):

• frameworks/sens-damage-categories.md — line 224 (Revel bullet) updated with reproducibility-setback narrative + scope-broadening to 5-or-6 target portfolio; line 225 (Alagebrium bullet) corrected “HFpEF/diastolic” to “systolic” (BENEFICIAL enrolled systolic HF, not HFpEF — this was an outright factual error in the prior page) + added Yang 2003 cross-reference; Draghici 2015 PMID corrected 26228140 → 26472902 (the prior value was wrong by ~1.5M PMIDs)
• processes/advanced-glycation-end-products.md — Yang 2003 footnote upgraded; body section expanded with three-compound framing; pyridoxamine subcategory caveat added

Frontmatter housekeeping:

• processes/glucosepane.md literature-checked-through: 2026-05-09 → 2026-05-15; verified-scope extended to flag 2026-05-15 partial additions as UNVERIFIED industry-watcher-grade for Revel content (#gap/needs-primary-source-verification)
• processes/advanced-glycation-end-products.md literature-checked-through: 2026-05-09 → 2026-05-15; banner updated to reflect 2026-05-15 web-research additions

Remaining verification priorities:

• Yang 2003 full-PDF verification — paper is closed-access (oa_status: closed in a local paper archive); ~$30 paywall via Elsevier or institutional access required. Abstract-level verification confirms the headline claim verbatim; PDF would add Tables 1–2, control conditions, and the dose-response (or lack thereof) of compound concentrations tested
• 12 alagebrium-class preclinical citations (Vasan/Wolffenbuttel/Liu/Candido/Chung/Park/Joshi/Pandhi/Cooper/Tang/Hu — abstract-level only) remain at “verification pending” status; lower priority than Yang 2003 since they are not the load-bearing critique
• Revel reproducibility-setback claim is sourced to a Fight Aging blog commenter (#gap/needs-primary-source-verification) — confirm via Revel official disclosure or peer-reviewed methodology paper before treating as established fact

[2026-05-15] update | processes/advanced-glycation-end-products.md + processes/glucosepane.md — AGE-breaker preclinical landscape

Triggered by user question about preclinical glucosepane-breaker candidates with published in-vivo data. Initial wiki scan found the topic was thinly covered: alagebrium framed as a Phase 2/3 clinical story without the preclinical record; the Spiegel/Revel program described as “preclinical chemistry and tool-antibody development.” Web research surfaced three substantive gaps in wiki coverage:

1. Alagebrium-class preclinical in-vivo data is extensive and was missing. ALT-711 has published in-vivo functional readouts in aged rhesus monkeys (Vasan/Wagle PNAS 2000 PMID:10706607), STZ-diabetic rats (Wolffenbuttel PNAS 1998), aged dogs (Asif/Kass series), aged diabetic rat heart (Liu 2003 PMID:12946933), diabetic rat myocardium (Candido Circ Res 2003 doi:10.1161/01.res.0000065620.39919.20), diabetic-hypertensive rats + nifedipine (Chung Hypertens Res 2014), and isolated rat carotid (Park 2017 PMC5665628). TRC4186 (Torrent Pharmaceuticals) has Ob-ZSF1 type-2 diabetic rat data (Joshi J Cardiovasc Pharmacol 2009 PMID:19546815) plus a completed Phase 1 clinical (Pandhi Clin Drug Investig 2009 doi:10.2165/11315260-000000000-00000) — program subsequently stalled. PTB (alagebrium’s precursor) has STZ-diabetic rat (Cooper 2000 PMID:10855541), bone fragility (Tang PLOS ONE 2014 doi:10.1371/journal.pone.0103199), and rat periodontitis (Hu 2014 PMID:24988129) papers.

2. Yang 2003 mechanism critique was missing. Yang/Litchfield/Baynes Arch Biochem Biophys 2003 PMID:12646266 showed that the AGE-breaker class cleaves in-vitro model compounds but does NOT cleave Maillard crosslinks formed in vivo in real-tissue skin/tail collagen from diabetic rats. This is a load-bearing critique of the entire alagebrium class — it implies the positive functional readouts above may reflect indirect mechanisms (RAGE modulation, carbonyl scavenging) rather than genuine ECM crosslink reversal. Should be a PDF-verification priority.

3. Revel Pharmaceuticals reproducibility setback and pivot was missing. Per industry-watcher reporting (Fight Aging 2021 article with 2024 commenter update; corroborated by Longevity.Technology + Lifespan.io coverage), the glucosepane-cleaving activity of the original Spiegel-lab enzyme candidates could not be reproduced, prompting Revel to broaden scope to a “5 or 6 target” portfolio of aging crosslinks rather than glucosepane exclusively. Strengthens the wiki’s needs-human-replication flag and contradicts the SENS-categories framing of “Highest-priority recent finding: Functional-metagenomics enzyme discovery (Revel, ~2023–2024) — represents the first credible glucosepane-degrading biocatalyst pipeline.”

• updated: processes/advanced-glycation-end-products.md

  • Frontmatter literature-checked-through: 2026-05-09 → 2026-05-15
  • Banner updated to flag the 2026-05-15 web-research additions
  • “Interventions — historical failures” section restructured:
    • Alagebrium subsection now includes the 7-row preclinical in-vivo table
    • New H3 subsection: “TRC4186 (Torrent Pharmaceuticals) — pyridinium AGE-breaker”
    • New H3 subsection: “N-phenacylthiazolium bromide (PTB) — the prototype α-dicarbonyl breaker”
    • New H3 subsection: “The Yang 2003 critique — do alagebrium-class breakers actually cleave real-tissue Maillard crosslinks?”
    • Conclusion rewritten to integrate the Yang 2003 critique and the Revel reproducibility setback
  • Revel Pharmaceuticals subsection updated with reproducibility-setback narrative + 5–6 target portfolio pivot
  • 13 new footnotes added: [^vasan2000] [^wolffenbuttel1998] [^liu2003] [^candido2003] [^chung2014] [^park2017] [^joshi2009] [^pandhi2009] [^cooper2000] [^tang2014] [^hu2014] [^yang2003] [

• updated: processes/glucosepane.md

  • Frontmatter literature-checked-through: 2026-05-09 → 2026-05-15
  • verified-scope extended to document the 2026-05-15 partial update is UNVERIFIED (industry-watcher-grade Revel content)
  • Revel/Spiegel section rewritten with reproducibility setback + pivot + explicit cross-reference to advanced-glycation-end-products for the alagebrium-class preclinical record and Yang 2003 critique
  • 1 new footnote added: [

• gaps surfaced:

  • #gap/needs-primary-source-verification on the Revel reproducibility-setback claim (sourced to a Fight Aging blog comment, not primary literature)
  • #gap/contradictory-evidence between the alagebrium-class positive in-vivo functional papers and the Yang 2003 ex-vivo collagen result
  • Yang 2003 (PMID:12646266) is a high-priority PDF-verification target — the foundational mechanism critique should be verified against full PDF before the wiki’s narrative leans on it heavily
  • 12 alagebrium-class preclinical citations added at abstract-level provenance — flagged as verification-pending in their respective footnotes

• downstream propagation needed:

  • frameworks/sens-damage-categories.md — “Highest-priority recent finding: Functional-metagenomics enzyme discovery (Revel, ~2023–2024)” framing should be softened in light of the reproducibility setback; the GlycoSENS Phase column (“Preclinical / None clinical / Revel Pharmaceuticals”) remains correct but the strategic-pivot context should be noted

[2026-05-14] verify | frameworks/interventions-testing-program.md

• page: frameworks/interventions-testing-program.md (type: framework; no verified flag)
• PDFs read in full: Nadon 2008, Cheng 2019, Jiang 2024, Snyder 2025
• abstracts only (not OA): Miller 2007, Nadon 2006, Korstanje 2026, Jiang 2025, Bou Sleiman 2022, Mokalla 2026; Parker 2025 (preprint)
• corrections applied:
  • Cheng 2019 “~25% female-survival advantage” → “~10% median lifespan advantage” (females 887 d vs males 803 d; ~10.5%)
  • Korstanje 2026 “8 compounds” → “11 interventions total” (8 named in title + 3 unnamed; abstract explicitly says eleven; 3 remaining unidentifiable without full text)
  • Snyder “2026” → confirmed as Snyder 2025 (GeroScience (2025) 48:1787–1797, online July 2025); corrected in footnote
  • Snyder footnote: paper covers BOTH 16α-hydroxyestradiol AND canagliflozin; now updated
  • Cohort sizes table: “~150–200/sex/site” → Nadon 2008 planned minimum is 36F+44M/site/arm; gap tag added
  • Green tea extract negative table entry: clarified GTE is positive by Gehan test in females (+7% p=0.0344 per Jiang 2024); row split from curcumin
  • Korstanje: added that pioglitazone and mitoglitazone reduced female lifespan at two sites (per abstract)
  • All 11 footnotes updated with quantitative anchors; verified-scope qualifiers updated per actual verification depth
• gaps surfaced: Korstanje 2026 full compound list (11 vs 8 discrepancy); not OA. Miller 2007 PDF not OA. Nadon 2006 not OA.
• downstream propagation needed: none — this is a framework page; compound pages (rapamycin, canagliflozin, metformin, acarbose, GTE) may benefit from the corrected Jiang 2024 GTE finding

[2026-05-14] verify | model-organisms/caenorhabditis-elegans.md — worm-specific confounds section (three new citations)

Pages verified: 1 (partial — three new 2026-05-14 citations only)

• model-organisms/caenorhabditis-elegans.md — corrections applied to “Worm-specific experimental confounds” section:
  • Cabreiro 2013 Cell (10.1016/j.cell.2013.02.035): verified against local PDF. Corrections: (1) “microbiome-mediated, not direct” → “primarily microbiome-mediated” (paper explicitly states both direct and indirect effects); (2) added dose-response quantitative anchors (+18%/+36%/+3% at 25/50/100 mM); (3) added axenic/UV-irradiated null finding; (4) added R²=0.82 bacterial-growth correlation; (5) added explicit dual direct+indirect framing from Discussion
  • Cabreiro & Gems 2013 EMBO Mol Med (10.1002/emmm.201100972): verified against local PDF (PMC3799487). Framing as “confounds catalog” confirmed accurate; updated footnote to reflect “Bridge the Gap” review format, holobiont framing, and Table 1 DR-method summary
  • Anderson 2016 Mech Ageing Dev (10.1016/j.mad.2016.01.004): verified against PMC full text (PMC4789167; local PDF unavailable). Corrections: (1) added osmotic-stress directional quantification (−50% without FUdR, +20–25% with); (2) added DAF-16::GFP +30% finding; (3) corrected sirtuin claim — single sir-2.1 deletion has minimal effect; triple sir-2.1;sir-2.3;sir-2.4 mutant required for greatly reduced extension; added author explicit recommendation to avoid FUdR
• Frontmatter: verified-date updated 2026-05-04 → 2026-05-14; literature-checked-through: 2026-05-14 added; verified-scope updated
• Banner updated to list all verified + remaining unverified citations

Pages unverifiable (closed-access, unchanged from prior pass):

• Kenyon 1993, Vellai 2003, Melendez 2003 — no-fulltext-access remains

Downstream pages to check (not propagated here — for main agent):

• None identified: the worm-confounds section is new and not yet cited from other pages

[2026-05-14] follow-up pass | C. elegans + ITP + hyperfunction-synthesis + Garner/Garland

Follow-up to the translation-failure-of-aging-interventions hypothesis-page ingest (below) — completing the open follow-ups identified at end-of-pass.

• substantive expansion: model-organisms/caenorhabditis-elegans.md — added (1) Divergent-systems subsections for no ECM/skeletal aging, endocrine simplicity, and intestinal polyploidy vs mammalian stem-cell renewal; (2) new Worm-specific experimental confounds section parallel to mouse lab-confound — OP50/HT115 bacterial-food confound (Cabreiro 2013 Cell metformin-microbiome demonstration), FUdR sterilization confound (Anderson 2016 Mech Ageing Dev showing FUdR interacts with FOXO/sirtuins/BER), plate-vs-liquid, no-PK, temperature × intervention, N2-strain artifacts; (3) Hallmarks of aging coverage table mapping the 12 hallmarks to worm-modelability (yes/partial/no) — telomere/senescence/inflammaging all “no”; proteostasis/autophagy/nutrient-sensing all “yes”; (4) For personal/home research relevance section — operational rules for what worm assays can/cannot tell you about your own aging; (5) cross-link to translation-failure-of-aging-interventions

• new primary sources (PubMed-efetch-metadata only):

  • Cabreiro 2013 Cell doi:10.1016/j.cell.2013.02.035 PMID 23540700 — metformin-microbiome-folate
  • Cabreiro & Gems 2013 EMBO Mol Med doi:10.1002/emmm.201100972 PMID 23913848 — microbiota-worm-aging review
  • Anderson 2016 Mech Ageing Dev doi:10.1016/j.mad.2016.01.004 PMID 26854551 — FUdR aging-pathway confound

• new framework MOC: frameworks/interventions-testing-program.md — type: framework page on NIA Interventions Testing Program. Sections: methodology (UM-HET3 / 3-site / both sexes / pre-specified), positive-results table, negative-results table (Korstanje 2026 8-compound failure cohort + earlier resveratrol/methylene-blue/curcumin/green-tea), sex-differences synthesis (Jiang 2025 + Bou Sleiman 2022 + Cheng 2019), what-ITP-validation-means-for-human-translation (necessary but not sufficient — lab-mouse confounds intact), cross-vertebrate corroboration (Ivimey-Cook 2025), open questions

• new primary sources cited (PubMed-efetch-metadata only):

  • Miller 2007 Aging Cell doi:10.1111/j.1474-9726.2007.00311.x — founding ITP design
  • Nadon 2008 Age doi:10.1007/s11357-008-9048-1 — full design rationale
  • Nadon 2006 Aging Cell doi:10.1111/j.1474-9726.2006.00185.x — pre-ITP rationale
  • Korstanje 2026 Geroscience PMID 41843349 — 8-compound negative cohort
  • Jiang 2025 J Gerontol A PMID 40717358 — sex-differences 2-decade review
  • Bou Sleiman 2022 Science PMID 36173858 — sex/age longevity QTLs
  • Cheng 2019 Aging Cell PMID 30801953 — female-survival-advantage in UM-HET3
  • Jiang 2024 Geroscience PMID 38630424 — Gehan-test methodology
  • Snyder 2026, Mokalla 2026, Parker 2025 — supporting

• synthesis development: hypotheses/hyperfunction-theory.md — added Notes/open-questions subsection “Hyperfunction × translation-failure: are already-aged trial cohorts the wrong test population?” developing the synthesis that hyperfunction-targeting interventions may underperform in aged human cohorts because hyperfunction-vs-damage shifts toward damage at human late-life ages; testable prediction = age-stratified mTOR-inhibitor RCT (45–55 vs 65–75 enrollment). Added translation-failure-of-aging-interventions to Related hypotheses section. Bidirectional link updated on translation-failure page (#gap/synthesis-needed marker resolved).

• second-pass Garner/Garland additions to model-organisms/mus-musculus.md:

  • Richter, Garner et al. 2011 PLoS ONE doi:10.1371/journal.pone.0016461 PMID 21305027 — population heterogenization (multi-strain) improves cross-lab reproducibility; counterexample to genetic-standardization-aids-reproducibility assumption
  • Swallow, Carter & Garland 1998 Behav Genet doi:10.1023/a:1021479331779 PMID 9670598 — founding paper of Garland HR (high-runner) mouse selection lines
  • Both PubMed-efetch-metadata only; verified-scope updated to flag these as not-yet-PDF-verified

• gaps resolved from previous ingest:

  • ✓ interventions-testing-program missing-page — created
  • ✓ hyperfunction × translation-failure synthesis — developed
  • ✓ Garner/Garland primary literature — integrated
  • ✓ Mattson/Rosshart/Runge/Ivimey-Cook verification — completed (entry below)

• gaps still open:

  • mouse-vs-human normalized-effect-size meta-analysis — no dedicated paper or wiki study page found
  • per-cohort study pages for ITP positive/negative results — substantial seeder work pending
  • verifier pass for the 12 new metadata-only citations across worm + ITP + mouse pages (Cabreiro/Anderson/Richter/Swallow/Miller-2007/Nadon-2008/Nadon-2006/Korstanje/Jiang-2025/Bou-Sleiman/Cheng/Jiang-2024)

[2026-05-14] verify | model-organisms/mus-musculus (4-citation PDF cross-check)

• page: model-organisms/mus-musculus.md — verified: true (all four 2026-05-14 additions now PDF-verified)
• sources checked: Martin/Mattson 2010 (doi:10.1073/pnas.0912955107), Rosshart 2017 (doi:10.1016/j.cell.2017.09.016), Runge 2025 (doi:10.1038/s41467-025-60554-2), Ivimey-Cook 2025 (doi:10.1111/acel.70131) — all four PDFs downloaded and read in full
• corrections made:
  • Rosshart 2017 body framing: “improved disease resistance and immune competence” → “improved disease resistance and reduced immunopathological inflammation” (the paper demonstrates limited cytokine-storm response and improved survival, not generalized immune enhancement; WildR mice actually show lower pro-inflammatory cytokines)
  • Rosshart 2017 footnote: added survival figures (92% WildR vs ~17% Lab/LabR, p<0.0001 log-rank, Fig. 5A); corrected model description to GF C57BL/6; added AOM/DSS tumorigenesis result; clarified mechanism
  • Runge 2025 footnote: added article number (16:5301); added key quantitative finding (2 genes TXwilding vs wildling, 1422 vs lab mice); noted scalable TX system
  • Ivimey-Cook 2025 footnote: added article number (24:e70131); added 911 effect sizes / 167 papers / 8 species; added p-values; added publication-bias caveat and I²=96.5% heterogeneity note
  • Martin 2010 footnote: corrected article type “review” → “perspective/review”; added Table 1 quantitative data (blood pressure, body weight ranges)
  • All four “verified-scope: PubMed efetch metadata only” qualifiers removed from footnotes; frontmatter verified-scope updated to reflect PDF verification
  • verified-date updated: 2026-05-04 → 2026-05-14
• no downstream propagation needed (corrections are to the mus-musculus page body and footnotes only; no other wiki page quotes these specific values)

[2026-05-14] ingest | hypotheses/translation-failure-of-aging-interventions

• added: hypotheses/translation-failure-of-aging-interventions.md — Mode B (conceptual-frame) hypothesis page synthesizing why aging interventions that look promising in model organisms produce null/modest results in humans; verified: false; literature-checked-through: 2026-05-14
• structure: thesis statement → case exhibits (CR, rapamycin, metformin, sirtuin/NAD class, senolytics) → three-tier over-determined cause framework (experimental-design asymmetry / species-biology divergence / selection bias and interpretation) → what-it-explains/doesn’t → tests-and-predictions-locations → related-frames → what-would-update-the-frame → practical-implications → limitations/gaps
• division of labor with model-organisms/_extrapolation-guide.md: extrapolation-guide is the DATA layer (per-organism divergence catalog, rubric, case pitfalls — “is this specific extrapolation safe?”); hypothesis page is the SYNTHESIS layer (cross-cutting “why does this pattern persist?”)
• updated: model-organisms/_extrapolation-guide.md — added synthesis-pointer paragraph near top linking to the new hypothesis page; division-of-labor explicit
• updated: model-organisms/mus-musculus.md — added “Lab-mouse confounds (design artifacts vs species biology)” section citing Martin/Mattson 2010 PNAS “metabolically morbid” canonical paper, Rosshart 2017 Cell wildling microbiome, Runge 2025 Nat Comms wildling phenotype; added Ivimey-Cook 2025 Aging Cell cross-vertebrate meta-analysis to CR failure-mode subsection (rapamycin/CR generalize across vertebrates, metformin does not); added cross-link to new hypothesis page in See also; extended verified-scope to note new citations are PubMed-efetch-metadata only (full texts NOT yet verified)
• new primary sources cited (PubMed efetch metadata only — not PDF-verified):
  • Martin B, Ji S, Maudsley S, Mattson MP 2010 PNAS doi:10.1073/pnas.0912955107 PMID 20194732 — canonical lab-rodent baseline-morbidity confound
  • Rosshart SP et al. 2017 Cell doi:10.1016/j.cell.2017.09.016 PMID 29056339 — wild-mouse microbiota engraftment improves fitness and disease resistance
  • Runge S et al. 2025 Nat Commun doi:10.1038/s41467-025-60554-2 PMID 40506454 — wildling phenotype follow-up
  • Ivimey-Cook ER, Sultanova Z, Maklakov AA 2025 Aging Cell doi:10.1111/acel.70131 PMID 40532901 — vertebrate meta-analysis: rapamycin + DR generalize, metformin does not
• updated (cross-link only — added new hypothesis page to “See also”): model-organisms/nothobranchius-furzeri.md, model-organisms/heterocephalus-glaber.md, model-organisms/rattus-norvegicus.md, model-organisms/homo-sapiens.md, model-organisms/canis-lupus-familiaris.md, model-organisms/drosophila-melanogaster.md
• not updated (no natural insertion point — no See also section): model-organisms/caenorhabditis-elegans.md, model-organisms/saccharomyces-cerevisiae.md; reachable via _extrapolation-guide.md backlink; model-organisms/balaena-mysticetus.md and model-organisms/loxodonta-africana.md skipped (comparative-biology not lab-model — translation-failure frame less directly applicable; reachable via mouse cross-links)
• gaps surfaced:
  • #gap/missing-page — interventions-testing-program.md does not exist; ITP is the most-cited mouse-aging methodology in the wiki; deserves dedicated entity page documenting design + translation-tracking table
  • #gap/synthesis-needed — interaction between hyperfunction-theory and translation-failure frame (does hyperfunction-vs-damage explain PEARL/RAPA-EX-01 nulls in aged cohorts?) is under-developed
  • #gap/needs-replication — systematic mouse-vs-human normalized-effect-size meta-analysis would directly test the central magnitude-shrinkage claim
  • #gap/no-fulltext-access — Ivimey-Cook 2025 lacks dedicated study page; verifier chain is one hop longer than ideal
  • #gap/unsourced — Garner (mouse housing welfare) and Garland (mouse exercise) primary citations not yet integrated; only Mattson covered in this pass

[2026-05-14] ingest | interventions/pharmacological/ppara-agonists

• added: interventions/pharmacological/ppara-agonists.md — PPARα agonists / fibrate class; type: intervention; verified: false
• updated: frameworks/intervention-classes.md — added ppara-agonism / cr-mimetic class entry (R16 discipline)
• literature-checked-through: 2026-05-14 (PubMed recency sweep 2021–2026; meta-analysis/RCT filter; 20+ results scanned; key recent: PROMINENT 2022 NEJM + Kim 2022 Metabolism)
• canonical IDs: no UniProt/PubChem for class page; ppara protein page is forward-reference stub
• clinical-trials-active: 0 (ClinicalTrials.gov v2 API: no active/recruiting trials with fibrate + aging/frailty/sarcopenia endpoints)
• primary sources confirmed via PubMed efetch + DOI lookup: FIELD 2005 (not_oa), ACCORD-Lipid 2010 (download pending), Helsinki 1987 (not_oa), Jo 2021 (download pending), Kim 2022 (download pending), Corton 2004 (download pending), Brandstädt 2013 (download pending), PROMINENT 2022 (download pending), McKenzie 2025 (locally available), Nikolai 2015 (download pending)
• R25 recency findings: PROMINENT 2022 (pemafibrate failed MACE; integrated as contradictory evidence vs lipid-lowering hypothesis); Kim 2022 (Metabolism cohort; HR 0.618 for >1 yr users); no recency hits contradicted training-era framing; no fibrate aging-endpoint RCTs found
• gaps surfaced: no mammalian lifespan study (#gap/needs-human-replication); no geroscience-endpoint RCT registered (#gap/needs-human-replication); LOY-002 API undisclosed (#gap/no-disclosure); PROMINENT vs observational-cohort discordance (#gap/contradictory-evidence); rodent neuroprotection claims unsourced (#gap/unsourced)
• implicit stubs created: ppara, fenofibrate, gemfibrozil, bezafibrate, ciprofibrate, pemafibrate

[2026-05-14] ingest | studies/coleman-2025-triad-protocol

• added: studies/coleman-2025-triad-protocol.md — TRIAD trial design/rationale (Coleman & Creevy et al. 2025); type: study; verified: false
• doi confirmed: 10.1007/s11357-024-01484-7 (matches PMC12181551 / PMID 39951177); local PDF available at a local paper archive
• DOI/PMID cross-check: PMC ID and DOI confirmed as consistent via PubMed efetch; lead co-first authors confirmed as Amanda E. Coleman and Kate E. Creevy
• co-first authors confirmed via PDF cover page: Coleman AE and Creevy KE contributed equally
• quantitative claims pulled from PDF: n=580 (290/arm); dose 0.15 mg/kg once weekly oral; eligibility ≥7 yr, 20–55 kg; 19 sites; trial status at writing 158 enrolled of 252 screened; funding grant UI9AG057377
• aliases set: [triad-trial, TRIAD, Test of Rapamycin in Aging Dogs] — resolves triad-trial wikilink from canis-lupus-familiaris.md
• implicit stubs created: urfer-2017-rapamycin-pilot, dog-aging-project (previously flagged as stub)
• gaps surfaced: NCT registration number not printed in PDF body (#gap/needs-nct-id-confirmation); NIH expansion grant amount claimed in public announcements not verifiable from paper (#gap/needs-citation-verification); trial results pending (#gap/long-term-unknown)
• literature-checked-through: 2026-05-14

[2026-05-14] ingest | studies/creevy-2022-dog-aging-project

• added: studies/creevy-2022-dog-aging-project.md — Dog Aging Project cohort/program anchor page; type: study; verified: false
• anchors: doi:10.1038/s41586-021-04282-9 (Creevy 2022 Nature); PMID 35110758; PMC8940555; local PDF confirmed at
• aliases: [dog-aging-project, DAP, Dog Aging Project] — resolves dog-aging-project wikilink from model-organisms/canis-lupus-familiaris.md
• literature-checked-through: 2026-05-14 (PubMed recency sweep; 20 DAP-related PMIDs triaged)
• primary sources verified via PubMed efetch + DOI lookup:
  • doi:10.1038/s41586-021-04282-9 (Creevy 2022; local PDF available)
  • doi:10.1007/s11357-025-01571-3 (Prescott 2025, Precision Cohort design; PMID 40038157; not_oa)
  • doi:10.1093/gerona/glaf279 (Harrison 2026, metabolomic mortality signature; PMID 41429575; download pending)
  • doi:10.1101/2024.10.03.616519 (McCoy 2024, epigenetic clock; PMID 39553930; preprint)
  • doi:10.1111/acel.70226 (Harrison 2025, protein catabolite biomarkers; PMID 41121965)
  • doi:10.1007/s11357-017-9972-z (Urfer 2017, rapamycin pilot; PMID 28374166; download pending)
  • doi:10.1007/s11357-024-01484-7 (TRIAD protocol; local PDF available)
  • doi:10.1371/journal.pcbi.1012728 (Fang 2025, comorbidity networks; PMID 40811246)
  • doi:10.2460/ajvr.25.02.0038 (Long 2025, CDI for multimorbidity; PMID 40164106)
  • doi:10.1002/alz.70630 (McGrath 2025, Alzheimer’s/CCD platform; PMID 40995864)
• implicit stubs created: triad-trial (cross-linked but may not yet exist)
• gaps surfaced: TRIAD results pending (#gap/long-term-unknown); Pack Cohort enrollment number unverified from peer-reviewed source (#gap/needs-citation-verification); Harrison 2026 metabolomics needs replication (#gap/needs-replication); multiple secondary publications abstract-only verified

[2026-05-14] verification | canis-lupus-familiaris

Pages verified: 1

• model-organisms/canis-lupus-familiaris.md — flipped to verified: true (partial scope — Sutter 2007 PDF unavailable)

Sources checked against local PDFs (7 papers):

• doi:10.1038/s41586-021-04282-9 (Creevy 2022 Nature) — confirmed: Kaeberlein at Princeton (Akey), cohort structure (5 nested cohorts), >30,000 owners at Dec 2020 snapshot; corrected enrollment description — “47,000” is post-publication figure, not stated in the paper
• doi:10.1007/s11357-024-01484-7 (Coleman/Creevy TRIAD 2025 GeroScience) — confirmed: N=580 randomized, 0.15 mg/kg once weekly, eligibility ≥7 yr + 20–55 kg + sterilized, 19 sites, 36-month duration; “first rigorous test” language confirmed verbatim in abstract
• doi:10.1086/669665 (Kraus 2013 Am Nat) — confirmed: 74 breeds, 56,637 individuals from VMDB, conclusion “large dogs die young mainly because they age quickly”
• doi:10.1007/s11357-017-9972-z (Urfer 2017 GeroScience) — confirmed: 29 randomized, 24 completed per-protocol; enriched footnote with 3-arm design, exact p-values for FS and E/A ratio, MCV finding
• doi:10.1186/s40575-020-00086-8 (Urfer 2020 Canine Med Genetics) — confirmed: n=20,970 (not ~21,000); gonadectomy effect sex-asymmetric (significant in females, not independently significant in males); body text updated to reflect sex-differential finding accurately
• doi:10.1093/gerona/glaf279 (Harrison 2026 J Gerontol A) — confirmed: n=937, published Dec 22 2025 (vol 81 issue 4 = 2026); enriched footnote with Pearson r=0.52 for human-dog HR correlation and 23 FDR≤5% metabolites
• doi:10.1007/s11357-023-00911-5 (Jiménez 2024 GeroScience) — confirmed: published Aug 2023 (journal vol 46 2024); covers exactly 10 hallmarks (list confirmed)

Sources checked via PubMed abstract (closed-access, full PDF unavailable):

• doi:10.1016/j.rvsc.2006.06.005 (Greer 2007) — PMID 16919689; abstract confirms 77 AKC breeds, n>700, r=−0.679 weight-lifespan, r=−0.603 height-lifespan; both p<0.05; values in wiki match abstract exactly

Sources not verifiable from full PDF:

• doi:10.1126/science.1137045 (Sutter 2007) — Science paywall; download failed; IGF1 haplotype claim and 669-citation count confirmed via DOI lookup; full PDF verification deferred; tagged in verified-scope

Corrections made:

1. Enrollment figure: “47,000 enrolled” → clarified as post-publication DAP communications; 2022 paper states >30,000; TRIAD 2025 paper states ~50,000 — body text updated with sourced figure
2. Urfer 2020 n: “n~21,000” → n=20,970 in body text; gonadectomy effect description corrected (significant in females, not independently significant in males after weight/clinic correction)
3. LOY-002 target population: “overweight middle-aged dogs (any size)” → “dogs ≥10 years weighing ≥14 lb” per Loyal press release
4. LOY-002 TAS date: “Dec 2025” → “Jan 13, 2026” per Loyal official blog post
5. TRIAD description: added confirmed N=580, dose 0.15 mg/kg once weekly, eligibility criteria, 36-month duration to body text and footnote
6. Urfer 2017 footnote: expanded from sparse description to include 3-arm design, exact n per group (29 randomized / 24 completed), specific statistical results, MCV finding
7. Creevy 2022 footnote: corrected Akey affiliation (Princeton, not U Washington); added cohort structure details; removed unverified enrollment figure
8. Harrison 2026 footnote: added Pearson r=0.52 and exact p-value; noted 104 deaths and mean 2.6-yr follow-up; added publication date
9. Jiménez 2024 footnote: added full list of 10 hallmarks covered; corrected to note published online Aug 2023

Loyal RXE/TAS dates verified against official press releases:

• LOY-001 RXE: November 28, 2023 — matches wiki “Nov 2023” ✓
• LOY-002 RXE: February 26, 2025 — matches wiki “Feb 2025” ✓
• LOY-002 TAS: January 13, 2026 — wiki had “Dec 2025” → corrected ✓

Supersession check (R25):

• Kraus 2013 (body-size-lifespan decomposition): PMID 37229711 (2023, Am Nat) is a related theoretical paper on evolutionary aging in dogs — confirms rather than supersedes Kraus 2013’s empirical decomposition; no meta-analysis or large-cohort replication found that contradicts
• Sutter 2007 (IGF1 haplotype): PMID 22903739 (IGF1R as secondary size determinant, 2012) builds on rather than supersedes Sutter; no newer GWAS superseding the single major IGF1 haplotype finding identified
• Greer 2007: closed-access; no evidence of retraction or correction identified
• Literature-checked-through: 2026-05-14 ✓

Downstream propagation needed:

• studies/creevy-2022-dog-aging-project.md — may still carry unverified enrollment figure; recommend updating
• studies/coleman-2025-triad-protocol.md — may lack the detailed eligibility and dosing confirmed above; recommend enriching
• Any page citing Urfer 2020 as “~21,000 dogs” should be updated to 20,970

[2026-05-14] ingest | model-organisms/canis-lupus-familiaris

• added: model-organisms/canis-lupus-familiaris.md — domestic dog (Canis lupus familiaris); type: model-organism; verified: false
• updated: model-organisms/README.md — added dog row to organism table
• updated: hallmarks/deregulated-nutrient-sensing.md — lightweight cross-link to dog page (breed-size IGF-1 natural experiment)
• literature-checked-through: 2026-05-14 (PubMed + Europe PMC recency sweep, 2024–2026)
• canonical IDs confirmed: NCBI Taxonomy 9615; genome assembly ROS_Cfam_1.0; AnAge max ~29 yr
• primary sources confirmed via PubMed efetch + DOI lookup:
  • doi:10.1038/s41586-021-04282-9 (Creevy 2022, DAP cohort; PDF locally available)
  • doi:10.1007/s11357-024-01484-7 (TRIAD protocol 2025; PDF locally available; PMC12181551)
  • doi:10.1126/science.1137045 (Sutter 2007, IGF1 haplotype; 669 citations; download pending)
  • doi:10.1086/669665 (Kraus 2013, size-lifespan decomposition; OA)
  • doi:10.1016/j.rvsc.2006.06.005 (Greer 2007, height/weight vs. lifespan; closed-access)
  • doi:10.1007/s11357-017-9972-z (Urfer 2017, rapamycin pilot n=24; 170 citations)
  • doi:10.1186/s40575-020-00086-8 (Urfer 2020, companion-dog lifespan ~21,000 dogs)
  • doi:10.1093/gerona/glaf279 (Harrison 2026, dogs/humans share metabolomic mortality signature; DAP Precision Cohort n=937)
  • doi:10.1007/s11357-023-00911-5 (Jiménez 2024, hallmarks of aging in dogs review; OA)
• gaps surfaced: Loyal API identity (#gap/no-disclosure); TRIAD lifespan results pending (#gap/long-term-unknown); within-dog IGF-1 mechanistic chain incomplete (#gap/needs-replication); Greer 2007 closed-access (#gap/no-fulltext-access)
• implicit stubs created: dog-aging-project, triad-trial, growth-hormone-receptor

[2026-05-13] verify | processes/mitochondrial-rna-leakage.md

• Page verified: processes/mitochondrial-rna-leakage.md (synthesis page; created 2026-05-13)
• Verification method: Cross-link integrity + quantitative consistency against three verified source pages (mavs.md, rig-i-mavs-pathway.md, type-i-interferon-signaling.md). Primary PDFs not re-read (sources verified during the 2026-05-13 type-I-IFN cluster sweep).
• Corrections applied: None — all cross-checks passed clean.
• Cross-reference integrations confirmed: sasp.md, cellular-senescence.md, mitochondrial-dysfunction.md, rig-i-mavs-pathway.md, mavs.md all contain the expected new cross-links.
• Flipped to verified: true; ⚠️ banner removed.

[2026-05-13] sweep | type-I IFN signaling cluster completion (12-page batch + schema R36)

User-directed “completionism” follow-on to the R34/R35 schema-gap-fill session. Two 6-page parallel seeder/verifier batches + one schema escalation (R36) + propagation pass. All 12 pages verified: true. Individual verifier-written log entries follow below.

R36 schema escalation — literature-checked-through: on type: protein (CLAUDE.md):

• 7+ pages already populated this field in practice (STING, IKKε, JAK1, TYK2, STAT1, MAVS, IFNAR2 — flagged by 4+ seeders during today’s sweep). Now formally documented as optional ISO date field; populate for proteins targeted by active clinical-drug-development pipelines or fast-moving aging-specific research. Lint-pass flag at 18 months (same cadence as type:pathway R35, type:biomarker R25, type:process R33).

New atomic pages (12; all verified: true):

Batch 1 (high aging-relevance): molecules/compounds/amlexanox.md, molecules/proteins/jak1.md, molecules/proteins/tyk2.md, molecules/proteins/stat1.md, molecules/proteins/mavs.md, molecules/proteins/ifnar2.md

Batch 2 (type-I IFN cluster completion): molecules/proteins/ifnar1.md, molecules/proteins/stat2.md, molecules/proteins/irf9.md, molecules/proteins/irf7.md, pathways/rig-i-mavs-pathway.md, pathways/tlr3-trif-pathway.md

(Detailed corrections per page are in the individual ## [2026-05-13] verify | … entries below; this entry covers cross-cutting patterns.)

Most consequential corrections (load-bearing claims caught by verifiers):

• amlexanox: 200 mg TID → actual 50 mg TID dose fabrication (Oral 2017 Fig 1A)
• JAK1: n per group inflated 2-2.5× (~20 → actual 8-9, Xu 2015); tofacitinib fabricated in mouse frailty experiment (only ruxolitinib used)
• TYK2: Armstrong 2025 mis-attributed as primary POETYK ITT (actually post-hoc apremilast-switcher analysis); selectivity precision (100-200× vs JAK1/3, 3000× vs JAK2)
• STAT1: Pang 2011 contains NO STAT1 data (myeloid bias paper in PNAS); Tierney 2014 is about STAT3 not STAT1 (paralog conflation; Nature Medicine); Gonzalo 2023 primary JAK inhibitor baricitinib not ruxolitinib; Stat1+/− extends lifespan 15%, complete KO detrimental
• MAVS: Xu 2005 DOI resolved (10.1016/j.molcel.2005.08.014, PMID 16153868); Victorelli 2025 mt-RNA-MAVS-SASP added during verification
• IFNAR2: active anifrolumab trial count 4 → actual 20 (live ClinicalTrials.gov v2 API); TULIP-2 P<0.001 → actual P=0.001
• IFNAR1: Hou 2022 DOI mismatch (acel.13582 resolves to unrelated Uhrlaub West-Nile paper); anifrolumab structural claim wrongly attributed to Stark & Darnell 2012 (predates 2021 drug approval by 9 years); Bhattacharya 2010 mechanism — ligand-induced degron kinase is TYK2-catalytic (not PKD); PERK does not directly phosphorylate IFNAR1
• STAT2: GOF mechanism — stabilize-active-STAT2 → actual loss-of-USP18-interaction at R148Q/R148W; Takahashi 2018 cell type IMR90 → actual TIG-3
• IRF9: Platanitis 2019 ARGUMENT INVERSION — wiki described classical ISGF3 assembly and attributed it to Platanitis, who actually CHALLENGES the classical model; Lei 2025 doesn’t mention IRF9 (seeder inferred from pathway logic); STAT2 Tyr689 → actual Tyr690
• IRF7: CRITICAL DIRECTION REVERSAL of Sugihara 2026 — pDC IFN-α actually DECLINES with frailty (not “tonic excess”); only general serum cytokines increase
• RIG-I/MAVS pathway: Seth 2005 peroxisome mislocalization mis-attributed (actually Dixit 2010); cell-line error (HEK293T+Vero → actual HEK293+5 lines); López-Polo 2024 Nat Commun identified as independent Serrano-lab replication
• TLR3/TRIF pathway: Srivastava 2025 wrong tissue (corneal → actual lung); wrong cytokine (IFN-β → actual IFN-α); Meylan 2004 PMID error (15467718 = unrelated mTORC2 paper; correct 15064760)

New seeder-fabrication sub-patterns observed today (reinforcing existing memory feedback_seeder_fabricates_outcomes.md + feedback_seeder_brief_doi_unreliable.md):

1. Cell-line/cell-type identity is unreliable in training memory. 4+ instances today (HUVEC↔IMR90/WI-38; TIG-3↔IMR90; HEK293T+Vero↔HEK293+5; corneal↔lung). Verifier brief should always include a “verify cell line/tissue against primary text” step.

2. Paper-conclusion direction can be inverted in training memory. 2 instances today (Sugihara 2026 frailty direction; Platanitis 2019 classical-vs-revised model). Verifier must read abstract minimum even for “confirmation” passes.

3. Paralog confusion — STAT1↔STAT3, JAK1↔JAK2 family-member identity is unreliable.

4. Anachronistic citation chains — Stark & Darnell 2012 cited for anifrolumab structural claim (drug approved 2021). Whenever citing a paper to support a specific drug or technology claim, check temporal feasibility.

Propagation pass results (all clean):

• ikbke.md inherited an identity-vs-similarity precision error from amlexanox verifier finding — corrected (65% similarity, not identity).
• mavs.md, type-I-IFN-signaling.md, ifnar2.md, jak-stat-pathway.md, irf3.md all cross-checked for inherited errors — clean.
• No other pages cite Bastard 2020, Jin 2022 ruxolitinib, Reilly 2017 dose, or any of today’s corrected DOIs.

R25 recency findings worth promoting beyond their seed pages:

• Victorelli 2025 + López-Polo 2024 + Zhang 2026: mtRNA-RIG-I/MDA5-MAVS-SASP story now reproduced across 3 independent labs (Passos, Serrano, Beijing groups) in 2 disease contexts (MASH + cognitive aging). Strong candidate for a dedicated section on [[mitochondrial-dysfunction]] or [[cellular-senescence]] hallmark pages.
• Bastard 2021 finer-band data: 80-85y prevalence peak + subsequent decline at >85y (~2.6%). Possible survivorship bias signal — anti-IFN-α2/IFN-ω carriers dying off before reaching the oldest-old cohort. Worth noting in a future MR/causal-inference review.
• Rentschler 2026 GeroScience — cytokine-induced senescence sustained by STAT1+NF-κB GRN signatures.

Implicit stubs surfaced (not seeded in this session; candidates for future):

• [[deucravacitinib]] compound (TYK2 inhibitor; Sotyktu)
• [[anifrolumab]] compound (anti-IFNAR1 mAb; Saphnelo) — should inherit verified TULIP-2 figures (86/180 vs 57/182; CI 6.3-26.3 pp; P=0.001; 20 active trials)
• [[rig-i]]/DDX58, [[mda5]]/IFIH1, [[lgp2]]/DHX58 — RIG-I/MAVS pathway nodes
• [[tlr3]], [[ticam1]]/[[trif]] — TLR3/TRIF pathway nodes
• [[usp18]] — IFNAR2/STAT2 feedback regulator
• [[bastard-2020-anti-ifn-covid-critical]] study page

Open gap items not resolved this session:

• IRF3 Ser386/Ser396 phosphorylation site citation (Fitzgerald 2003 + Meylan 2004 ruled out by verifier; likely Servant 2003 or Suhara 2000)
• Hou 2022 IFNAR1-in-aged-COVID paper — correct DOI not yet identified; tagged #gap/unsourced

[2026-05-09] verify — molecules/proteins/fpr2.md

Pages verified: 1 (partial scope)

Sources verified:

• Cooray 2013 PNAS (doi:10.1073/pnas.1308253110) — PDF downloaded and read end-to-end (6 pages)
• Bozinovski 2014 J Thorac Dis (doi:10.3978/j.issn.2072-1439.2014.08.08) — PDF downloaded and read end-to-end (5 pages)
• Slayo 2025 Mol Neurodegener (doi:10.1186/s13024-025-00824-1) — PDF downloaded and read (6 of 18 pages; abstract, intro, methods, key figures)
• Wang 2014/2015 Alzheimer’s & Dementia (doi:10.1016/j.jalz.2013.12.024) — verified via PMC full text (PMC4275415)

Unverifiable:

• Le 2002 Trends Immunol (doi:10.1016/s1471-4906(02)02316-5) — closed-access, not_oa
• Iribarren 2005 Immunol Res (doi:10.1385/ir:31:3:165) — closed-access, not_oa
• Claria 2012 J Immunol (doi:10.4049/jimmunol.1201272) — PDF download failed (3 attempts); partially verified from Crossref metadata + abstract

Corrections made:

• MEDIUM — Cooray 2013: SAA dimerization state in biased-agonism table wrong. Wiki said “Heterodimer-like”; paper shows SAA does NOT promote ALX homodimerization (decreases BRET signal); SAA’s pro-inflammatory signaling runs via monomers/FPR1-ALX heterodimers. Table corrected.
• MEDIUM — Cooray 2013: Mechanistic summary updated — AnxA1 IL-10 production confirmed in primary monocytes AND in vivo (mouse i.p. 1 µg AnxA1, absent in Alx-Fpr2/3 KO mice); FPR1/ALX heterodimer → JNK pathway added (paper’s Fig. 4 / neutrophil apoptosis data).
• MEDIUM — Cooray 2013: Footnote n= corrected from “n=cell-lines” to full model description: transfected HEK293, primary human monocytes, primary neutrophils, and C57BL/6 + Alx-Fpr2/3 KO mice. Full author list added (Clark AJL, Thompson D, McArthur S omitted in original).
• MAJOR — Wang 2014 AD section: Overstated “Elevated SPM receptor expression in AD brains (compensatory upregulation)” — FPR2/LXA4R protein was NOT significantly elevated by Western blot (p=0.958). Only IHC showed stronger staining. ChemR23 was significantly elevated (p=0.008). Corrected with specific statistics.
• MEDIUM — Wang 2014: Hippocampal RvD1 was not significantly reduced in AD (p=0.62) — this nuance was missing from the wiki, which implied RvD1 was also reduced. Corrected.
• MEDIUM — Wang 2014: Quantitative values added throughout (p-values, Spearman r, group n).
• MEDIUM — Wang 2014: Year in body text corrected from “2014” to “2015” (journal volume/issue) with online-publication caveat; footnote expanded with full author list and statistics.
• MINOR — Claria 2012: Author list corrected (2 authors → 5: added Dalli J, Yacoubian S, Gao F). Receptor attribution clarified: RvD2 acts via GPR32/ChemR23, NOT primarily via ALX/FPR2.
• MINOR — Bozinovski 2014: Second author “Anthony D” added (was omitted). Page range corrected: 1548–1558 → 1548–1556. Mitochondrial hexapeptides added as third pro-inflammatory FPR2 agonist class (per paper’s figure 1). SAA homodimerization noted.
• MINOR — Slayo 2025: Footnote expanded with full author list and specific mechanistic pathway (AMPK → SIRT1 → PGC-1α → oxidative phosphorylation); RvD4 unknown-receptor note added.
• Canonical IDs confirmed via UniProt REST API: P25090, NCBI Gene 2358, HGNC 3827, ENSG00000171049 all correct.
• Auto-extraction banner removed.

Downstream pages to check:

• None currently link to fpr2.md using fpr2 or … style citations that would need propagation. Page is a new seed with no known downstream dependents yet.

[2026-05-09] verify — biomarkers/vo2max-biomarker.md

Pages verified: 1 (partial scope — Mandsager 2018, Kokkinos 2010, Lang 2024, Kjaergaard 2025 verified against full PDF; Kodama 2009 and Poon 2021 closed-access not_oa; Dehn 1972 pre-DOI metadata confirmed via PubMed only)

Corrections made:

• MAJOR — Mandsager 2018 primary comparison reference group wrong: “Elite vs Below-Average HR 0.20” corrected to “Elite vs Low HR 0.20 (95% CI 0.16–0.24)”. Entire fabricated “HR vs Below-Average” table replaced with actual pairwise comparisons from Figure 2C.
• MAJOR — Tanaka 2000 citation mis-attributed: seeder cited JACC 2000;35(3):726-731 for VO₂max in masters athletes; actual Tanaka/Monahan/Seals JACC paper (doi:10.1016/s0735-1097(00)01054-8) is vol.37 (2001), pp.153-156, on age-predicted max HR — different topic. Footnote flagged as mis-attributed; Wilson & Tanaka 2000 AJP-Heart noted as possible intended source.
• Mandsager 2018: comorbidity HR comparisons now cite actual Figure 2C values (CAD HR 1.29, smoking 1.41, diabetes 1.40) instead of imprecise paraphrase.
• Kjaergaard 2025: journal year “2024” → “2025;110(5):1451-1459”; archive status updated from pending to PMC verified (PMC12012764).
• Kokkinos 2010: per-MET CI added (0.86-0.90); MET tier CIs added; person-years and study date range added; archive status updated to PDF verified.
• Lang 2024: full 11-author list added; per-MET HRs specified (0.89 and 0.83); heart failure HR 0.31 sourced to Aune 2021 within Lang.
• Dehn 1972: PMID 4643862 added; pre-DOI status and unverified quantitative claim noted.
• Kodama 2009 section: no-fulltext-access added; n discrepancy with Lang 2024 cited values flagged.
• Poon 2021: no-fulltext-access added to body and footnote.
• Auto-extraction banner removed.

Downstream pages to check:

• interventions/lifestyle/exercise.md — may cite Mandsager 2018 using “Elite vs Below-Average” framing (wrong reference group)
• Any page citing Tanaka 2000 JACC 35:3:726-731 for VO₂max/masters-athlete claims

[2026-05-09] verify — biomarkers/homa-ir-biomarker.md

Pages verified: 1 (partial scope — 3 of 5 cited primary sources verified against full PDF; Matthews 1985 PDF blocked; Weiss 2006 and Hwang 2015 closed-access not_oa)

Sources verified against full PDF:

• Wallace 2004 Diabetes Care (doi:10.2337/diacare.27.6.1487) — downloaded and read end-to-end
• Fontana 2010 AGE (doi:10.1007/s11357-009-9118-z) — downloaded and read end-to-end
• Rodriguez-Manas 2022 GeroScience (doi:10.1007/s11357-021-00384-4) — downloaded via PMC9135930; read end-to-end

Sources partially verified (PDF unavailable):

• Matthews 1985 Diabetologia (doi:10.1007/BF00280883) — Springer OA URL returns 303; formula and Rs=0.88 confirmed via Wallace 2004 restatement and PubMed abstract; tagged no-fulltext-access

Sources unverifiable (closed-access):

• Weiss 2006 Am J Clin Nutr (doi:10.1093/ajcn/84.5.1033) — not_oa
• Hwang 2015 Rejuvenation Research (doi:10.1089/rej.2015.1699) — not_oa

Corrections made:

• Removed auto-extraction banner
• Matthews 1985 validation n: “~20” → “n=23 (NGT=12, diabetes=11)” per Wallace 2004 Table 1; full paper n noted as unverified
• Matthews 1985 download status updated: pending → failed; no-fulltext-access tagged
• Wallace 2004 HOMA publication count: “500+” → “572” (exact figure stated in paper)
• Fontana 2010 mean CR duration: “~6 years” → “~7 years (range 3–20 years)”
• Fontana 2010 body text: added actual HOMA-IR values (CR 0.29±0.1 vs EX 0.44±0.3 vs WD 1.6±1.3, P=0.0001)
• Fontana 2010 intervention table: replaced vague “substantially reduced” with exact group values; labeled as cross-sectional comparison not a reduction endpoint
• Rodriguez-Manas 2022 footnote: expanded with full HR CIs across all 4 models; FTS OR=1.28 (1.01-1.63) added; Kaplan-Meier p=0.0082 added
• Metformin ~20-30% reduction: misattribution to Wallace 2004 removed; unsourced tagged
• Aging trajectory paragraph: Wallace 2004 citation removed; unsourced tagged
• Cutoff thresholds: explicit note added that Matthews 1985 and Wallace 2004 do not state categorical thresholds; cutoffs-not-from-primary-source tagged

Unverifiable claims:

• Weiss 2006 outcomes and Hwang 2015 OR — not verified (closed-access)
• Rapamycin and MK-677 HOMA-IR adverse signals — no primary footnote citations; noted as documented in transplant/trial literature but specific RCTs not cited

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

Pages verified: 1 (partial scope — Howard 1996, Barzilai 2012, Picca 2022 verified against local PDF; Sun 2020 verified via PubMed abstract + PMC; Ramirez 2019 not_oa; Pradhan 2013 download failed; canonical IDs confirmed via UniProt REST)

Corrections made:

• MAJOR — extrapolation table “96% identity human/rat” → “~93% identical, ~98% similar at amino acid level, human vs swine” (Howard 1996 reports human/swine only; rat never mentioned; 96% figure absent from paper)
• MAJOR — Picca 2022 claim overstated: paper actually states plasma ghrelin and acyl-ghrelin do NOT vary consistently with aging; corrected to reflect review framing of multiple candidates
• Constitutive activity primary source: [^holst2003] added (Holst 2003 Mol Endocrinol doi:10.1210/me.2003-0069); “~50% of maximum” figure removed (not confirmable from available sources)
• LEAP2 section: [^ge2018] added (Ge 2018 Cell Metab doi:10.1016/j.cmet.2017.10.016) as primary discovery citation
• Ramirez 2019 footnote: status corrected to not_oa with no-fulltext-access
• Pradhan 2013 footnote: status updated to failed download with no-fulltext-access
• Barzilai 2012 footnote: clarified as review; primary dwarf-mouse lifespan data from cited Brown 2001/Coschigano 2003
• Banner removed; verified: true (partial scope), 2026-05-09, claude

Downstream pages to check:

• molecules/compounds/mk-677.md — Barzilai 2012 dwarf mouse lifespan % claims may need same caveat
• hallmarks/deregulated-nutrient-sensing.md — may cite GHSR constitutive activity or Picca 2022 anorexia-of-aging framing

[2026-05-09] verify — biomarkers/grip-strength-biomarker.md

Pages verified: 1 (partial scope — 5 of 6 cited sources verified against full PDF; Park 2022 closed-access not_oa)

Sources verified against local PDF:

• Cruz-Jentoft 2019 EWGSOP2 (doi:10.1093/ageing/afy169) — pre-existing local PDF; Table 3 cutoffs confirmed; peak age corrected
• Granic 2016 Newcastle 85+ (doi:10.1371/journal.pone.0163183) — downloaded; n corrected in frontmatter; SD→SE corrected in footnote for −1.13 figure
• Ling 2010 Leiden 85-plus (doi:10.1503/cmaj.091278) — downloaded; HRs confirmed; p-values and follow-up range added to footnote
• Celis-Morales 2018 UK Biobank BMJ (doi:10.1136/bmj.k1651) — downloaded; age range and follow-up added; FNIH vs EWGSOP2 cutoff distinction added
• Bohannon 2019 NHANES (doi:10.5014/ajot.2019.029538) — downloaded; article number corrected (060→080); protocol-mismatch caveat added

Sources unverifiable:

• Park 2022 Am J Cardiol (doi:10.1016/j.amjcard.2021.08.061) — not_oa; endpoint corrected from “CAD” to “MI/cardiovascular fatality” per archive title metadata; tagged no-fulltext-access

Corrections: Peak age 30–35 → ~40 (Cruz-Jentoft); Granic n “not stated” → 845; Granic SD→SE; Bohannon article ID 060→080; Park endpoint CAD→MI; Celis-Morales age/follow-up added; Bohannon NHANES caveats added

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

Pages verified: 1 (partial scope — 6 of 8 primary sources verified; 2 closed-access)

Sources verified against full text:

• Thevaranjan 2017 Cell Host Microbe (doi:10.1016/j.chom.2017.03.002) — local PDF, read end-to-end
• Cani 2007 Diabetes (doi:10.2337/db06-1491) — downloaded (bronze OA via institutional repository), read methods+results
• Zhao 2023 J Neuroinflammation (doi:10.1186/s12974-023-02846-2) — downloaded (gold OA / PMC), read end-to-end
• Milan 2017 Nutrients (doi:10.3390/nu9040354) — downloaded (PMC), read end-to-end
• Ghosh 2015 J Gerontol (doi:10.1093/gerona/glu067) — verified via PMC full-text (PMC4311182; local download failed)
• Roberts 2020 Aging Clin Exp Res (doi:10.1007/s40520-020-01684-z) — verified via PMC full-text (PMC7937758; local download failed)

Sources unverifiable (closed-access):

• Moreno-Navarrete 2012 Int J Obes (doi:10.1038/ijo.2011.256) — not_oa; tagged no-fulltext-access in footnote
• Lopez-Moreno 2018 Exp Gerontol (doi:10.1016/j.exger.2017.11.006) — not_oa; tagged no-fulltext-access in footnote

Corrections made:

• Removed ⚠️ auto-extraction banner
• Thevaranjan aged mouse age: “19–23 months” → “18–22 months” (paper used 10–14 wk young vs 18–22 months old for macrophage experiments; FITC-dextran permeability measured at 3/12/15/18 months); n values added to footnote
• Ghosh 2015 exercise duration: “4-month” → “16-week” throughout (source’s exact language); n values added (n=12 older, n=13 young); quantitative effect sizes added to text and footnote (LPS 1.8-fold, LBP 1.9-fold, muscle TLR4 protein 2.1-fold, mRNA 3.8-fold; NF-κB p65 1.6-fold, p50 1.5-fold; exercise improved insulin sensitivity ~21%)
• Roberts 2020 direction corrected: wiki implied positive LBP–CVD risk association; actual finding is negative (r = −0.462, p<0.0001 — higher Framingham CHD risk correlated with LOWER LBP); n added (74); year corrected from 2021 → 2020; body text section rewritten to accurately represent the unexpected inverse direction
• Cani 2007: HFD LPS fold-change corrected from “2–3×” → “2.7-fold” (72% HFD arm); n values and CD14 KO strain detail added to footnote
• Milan 2017: n=15/group added to footnote and footnote redesign corrected (cross-over RCT, not observational); archive status updated to “local PDF available”
• Reference range table: “Older adults (70+) ~20–30 — Ghosh 2015; Roberts 2020” corrected — Ghosh 2015 reports fold-change not absolute; Roberts 2020 does not provide reference-range data; unsourced added
• Intervention table: “Aerobic exercise (4-month)” → “Aerobic exercise (16-week)”

Unverifiable claims (closed-access papers):

• Moreno-Navarrete 2012: n=222/34, LBP 22.2→16.2 μg/mL post-surgery — not verified against full PDF
• Lopez-Moreno 2018: exact dietary LBP/LPS effects, n, and LBP-specific results — not verified against full PDF

Downstream pages potentially affected:

• studies/thevaranjan-2017-aging-dysbiosis-permeability.md — if it exists, aged mouse age field may need updating
• processes/gut-barrier.md — cites Thevaranjan 2017; aged mouse age may propagate
• microbiome/gut-microbiome-aging-shifts.md — may cite Thevaranjan 2017

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

Pages verified: 1 (partial scope — 4 of 6 primary sources verified against PDF; Brzoska 2008 unverifiable; Brzoska 2010 book chapter not_oa)

Sources verified against PDF:

• Dalmasso 2008 Gastroenterology (doi:10.1053/j.gastro.2007.10.026) — local PDF confirmed
• Catania 2010 ScientificWorldJournal (doi:10.1100/tsw.2010.173) — downloaded (PMC) and confirmed
• Rinne 2013 Cardiovasc Res (doi:10.1093/cvr/cvs335) — downloaded (Europe PMC) and confirmed
• Sasaki 2014 Diabetologia (doi:10.1007/s00125-013-3140-5) — downloaded (Springer OA) and confirmed

Sources unverifiable:

• Brzoska 2008 Endocrine Reviews (doi:10.1210/er.2007-0027) — bronze OA URL blocked by Cloudflare; download failed; tagged no-fulltext-access
• Brzoska 2010 Adv Exp Med Biol chapter (doi:10.1007/978-1-4419-6354-3_8) — confirmed not_oa; KPV receptor-independence claims corroborated by Dalmasso 2008 primary data

UniProt P01189 fields verified via REST API.

Corrections made:

• β-MSH residue positions: 189–206 → 217–234 (UniProt P01189 confirmed)
• Dalmasso 2008 footnote n: “n≈40 mice per group” → “n=10 mice/group” (DSS model; TNBS model also n=10/group); mouse strain added (female C57BL/6, 8 wk); KPV Km value (≈160 μmol/L) added
• Sasaki 2014 body text and footnote: mechanism clarified (energy expenditure, not food intake, was primary driver in POMC-neuron model); male-specificity of phenotype noted; mouse background corrected (mixed 129/J × C57BL/6J, not pure C57BL/6)
• Rinne 2013 body text expanded: eNOS Ser1177 phosphorylation + Mn-SOD upregulation mechanism added; specific model details added (C57BL/6N acute, DIO chronic, MSH-OE transgenic)
• Anti-inflammatory section: IRAK-1/IRAK-M mechanism (from Catania 2010) added; JAK/STAT via MC5R + SOCS3 induction added; ERK1/2 attenuation in IECs added

Downstream pages potentially affected:

• molecules/compounds/kpv.md — cites Dalmasso 2008; n correction and Km value may propagate
• molecules/compounds/melanotan-ii.md — cites Rinne 2013 indirectly; no direct claim affected

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

Pages verified: 1 (partial scope — 3 of 5 primary sources verified against PDF; 1 not_oa; 1 not yet in archive)

Sources verified against PDF:

• Khalil 2020 Nat Commun (doi:10.1038/s41467-020-14612-6) — local PDF confirmed
• Disanto 2017 Ann Neurol (doi:10.1002/ana.24954) — downloaded (PMC) and confirmed
• Mattsson 2019 JAMA Neurol (doi:10.1001/jamaneurol.2019.0765) — downloaded (PMC) and confirmed

Sources unverifiable:

• Benkert 2022 Lancet Neurol (doi:10.1016/S1474-4422(22)00009-6) — not_oa; tagged no-fulltext-access
• Baskaran 2026 JAMA Cardiol (doi:10.1001/jamacardio.2026.0922) — not yet in archive; Crossref abstract-only cross-check

Corrections made (6 major, several minor):

• Khalil 2020 longitudinal n: “n=103” → “n=95 final dataset” (103 agreed; 8 excluded for incident disease)
• Khalil 2020 age-trajectory table: “20–50 yr” lower bound → “~40–60 yr”; added per-decade annualized rates (60–70: 4.32%/yr; >70: 4.23%/yr) from Table 3
• Disanto 2017 n: “n=219 MS + 40 controls” → Lugano n=142 MS + 254 HC (cross-sectional); SMSC n=246 MS (longitudinal) — MAJOR; prior values matched no group in the paper
• Disanto 2017 serum-CSF correlation: “r=0.86” → “Pearson r=0.77 (95% CI 0.69–0.83)” — MAJOR; 0.86 appears nowhere in the paper
• Mattsson 2019 cohort: “BioFINDER” → “ADNI” — MAJOR wrong cohort name
• Mattsson 2019 n: “1,182” → “1,583” — MAJOR
• Mattsson 2019 “NfL increases in AD patients but not in cognitively normal controls” → “NfL increases in all diagnostic groups; rates differ (CU 2.4, MCI 2.7, AD 4.9 ng/L/yr)” — MAJOR; claim that controls don’t increase was wrong
• Mattsson 2019 “Aβ-PET burden” → “CSF Aβ42 and FDG-PET” (paper used CSF Aβ42 as primary Aβ classifier)
• Baskaran 2026 aHR: “1.41 for all-cause mortality” → “1.35 for major vascular events (95% CI 1.22–1.50)”; individual mortality/CV-death aHRs flagged unverified
• FDA 510(k) gap note expanded: clearance year ~2022 from press-record evidence; K-number remains unconfirmed

Downstream pages that may need updates: none identified (nfl-biomarker.md has no inbound citations from entity pages yet)

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

Pages verified: 1 (partial scope — 5 of 6 sources verified against PDF; 1 closed-access)

Sources verified against PDF: Erickson 2011 (PNAS), Wrann 2013 (Cell Metab), Egan 2003 (Cell), Sleiman 2016 (eLife), Hartmann 2001 (EMBO J). Sources unverifiable: Connor 1997 (Mol Brain Res) — not_oa; tagged no-fulltext-access.

Corrections made:

• MAJOR — Erickson 2011 memory result: “spatial memory improved in aerobic group” implied between-group superiority → corrected; Time×Group for memory p=0.40 (ns); both groups improved equally.
• MAJOR — Erickson 2011 BDNF framing: corrected; Time×Group for serum BDNF p=0.23 (ns); BDNF-volume correlation was within-aerobic-group only (r=0.36–0.37, p<0.01).
• Erickson 2011 per-group ages and exact volume changes added; unsourced “150 min/week” protocol removed, actual trial protocol substituted.
• Wrann 2013 irisin caveat added: irisin did not significantly increase Bdnf in primary hippocampal neurons (Figure 7D); mouse strain corrected to “C57/Bl6” (no J suffix); BALB/c used for peripheral experiment; global (not muscle-specific) Pgc1a KO specified.
• Egan 2003 fMRI task: “memory encoding” → “N-back working memory task”; WMS-R measure specified; secretion values added (~19 vs ~30 ng/ml).
• Sleiman 2016: DBHB nomenclature added; n values added throughout; DBHB transporter-inhibitor rescue added; HDAC1 unaffected noted.
• Hartmann 2001 footnote: exact stimulus parameters, rat embryonic day, n per experiment added.
• Connor 1997 footnote: no-fulltext-access tagged.
• Banner removed; verified: true (partial scope), 2026-05-09, claude.

Downstream pages to check:

• exercise — Erickson 2011 memory/BDNF between-group claims
• altered-intercellular-communication — Connor 1997 AD claims
• stem-cell-exhaustion — Wrann 2013 irisin caveat if cited

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

Pages verified: 1 (partial scope — 6 of 7 sources verified against PDF; 1 closed-access)

Sources verified against PDF:

• Fest 2018 Eur J Epidemiol (doi:10.1007/s10654-018-0472-y) — downloaded and confirmed
• Fest 2018 Sci Reports (doi:10.1038/s41598-018-28646-w) — downloaded and confirmed
• Adamstein 2021 Eur Heart J (doi:10.1093/eurheartj/ehaa1034) — downloaded and confirmed
• Welsh 2018 ATVB (doi:10.1161/ATVBAHA.118.310945) — downloaded and confirmed
• Liu 2024 BMC Public Health (doi:10.1186/s12889-024-18888-x) — downloaded and confirmed
• Siedlinski 2020 Circulation (doi:10.1161/CIRCULATIONAHA.119.045102) — downloaded and confirmed

Sources unverifiable:

• Templeton 2014 JNCI (doi:10.1093/jnci/dju124) — not_oa; HR 1.81 and 100-study count tagged no-fulltext-access

Corrections made:

• Fest 2018 Eur J Epidemiol n: 8,711 → 8,715
• Fest 2018 Eur J Epidemiol follow-up: “median 7 years” → “mean 7.7 years (max 11.7 years)”
• Fest 2018 Eur J Epidemiol HR 1.64 attribution: “top vs bottom NLR quartile” → “per log-NLR unit continuous model”; Q4 vs Q1 HR 1.59 (1.37–1.86) now also documented
• Fest 2018 Sci Reports n: 4,650 → 8,711 (MAJOR — 4,650 appears nowhere in the paper)
• Fest 2018 Sci Reports “IQR 0.83–3.92” → “2.5th–97.5th percentile reference interval 0.83–3.92” (not IQR)
• Aging trajectory table: replaced with actual Fest 2018 Table 2 age-stratified means; removed unsourced below-age-45 estimates
• Adamstein 2021 n: “>10,061” → “60,087 total across 5 trials” (10,061 was CANTOS-only)
• Adamstein 2021 MACE HR range “1.09–1.31” → replaced with CANTOS-specific HRadj Q4 vs Q1 = 1.51 (1.30–1.75) and per-quartile effect 1.14 (1.09–1.19)
• Adamstein 2021 “statins did not lower NLR” → “lipid-lowering agents (rosuvastatin, bococizumab) had no NLR effect”; methotrexate increased NLR noted
• Welsh 2018 n: 393,368 → 478,259 (MAJOR — 393,368 appears nowhere in the paper)
• Welsh 2018 “lymphocyte count inversely associated” → “broadly flat/non-monotonic for CVD mortality”
• Welsh 2018 sex-specific HRs added: men 1.59 (1.22–2.08), women 2.15 (1.38–3.35) CVD mortality; women higher than men
• Liu 2024 n: 387,739 → 448,574 (MAJOR — 387,739 appears nowhere in the paper)
• Liu 2024 age range: “38–73” → “40–70”
• Liu 2024 life expectancy: “~3–4 yr” → sex-stratified: men 4.14 yr (3.75–4.56), women 2.96 yr (2.53–3.41) at age 60
• Siedlinski 2020: causal estimates added (SBP 0.69 mmHg/SD, DBP 0.56 mmHg/SD); n corrected (384,721 observational); no neutrophil MR signal noted
• All footnotes updated with correct n, journal/page, DOI lookup status

Downstream pages that may need updates:

• None identified; NLR biomarker page has no current inbound study links from entity pages

[2026-05-09] verify — biomarkers/il-6-biomarker.md

Pages verified: 1 (partial scope — 4 of 8 sources verified against PDF; 2 closed-access; 1 download failed; 1 not checked)

Sources verified against PDF:

• Sayed 2021 (Nature Aging, iAge) — local PDF confirmed
• Ridker 2017 CANTOS (NEJM) — local PDF confirmed
• Swerdlow 2012 IL6R MR Consortium (Lancet) — downloaded and confirmed
• Cesari 2003 Health ABC (Circulation) — downloaded and confirmed

Sources unverifiable:

• Harris 1999 EPESE (Am J Med) — not_oa; tagged no-fulltext-access in footnote
• Cesari 2004 InCHIANTI (J Gerontol) — download failed; not_oa per Europe PMC (PMID 15031308); tagged no-fulltext-access
• Ferrucci 2002 WHAS (JAGS) — not attempted (pending)
• Mathur 2008, Traise 2024 — not checked

Corrections made:

• iAge / IL-6 role (MAJOR): “IL-6 remains one of the strongest individual predictors within such composite models” → REMOVED and replaced with accurate statement: the Sayed 2021 paper explicitly states IL-6 and TNF-α were NOT major contributors to iAge; the top-15 Jacobians did not include IL-6. This was a direct contradiction of the source.
• Sayed 2021 marker count: “50+” → “exactly 50 cytokines, chemokines and growth factors” (paper states this precisely)
• Health ABC analysis n: “3,075” (enrolled) → “2,225 analyzed (3,075 enrolled)” throughout body text, cohort summary table, and frontmatter training-cohort field. Paper excluded 816 with baseline CVD and 34 with missing data.
• Swerdlow 2012 OR: “OR ~0.95 per allele” → “OR 0.95 (95% CI 0.93–0.97), p=1.53×10⁻⁵”; body text and footnote enriched with exact values and mechanism note (IL-6 rises with the minor allele because less is consumed by receptor)
• CANTOS footnote: enriched with full HR/CI (150 mg: HR 0.85, 95% CI 0.74–0.98, P=0.021), key secondary endpoint result, and clarification on triglyceride increase
• Sayed 2021 footnote: replaced “IL-6 among key predictors” with accurate “IL-6 and TNF-α were NOT major contributors to iAge”; added full top-15 Jacobian list and Framingham validation details
• Auto-extraction banner removed
• verified flag flipped to true with partial scope (Harris 1999 + Cesari 2004 body claims unverified)

Downstream pages that may need updates:

• chronic-inflammation — may reference IL-6 as a strong iAge contributor; needs correction if so
• Any page citing il-6-biomarker for the claim that “IL-6 is among the strongest individual predictors within composite inflammatory aging models”

[2026-05-09] verify — biomarkers/igf-1-biomarker.md

Pages verified: 1

• biomarkers/igf-1-biomarker.md (partial scope) — 5 primary source PDFs read end-to-end (Milman 2014, Fontana 2016 CALERIE, Watts 2022/2023, Vitale 2012, Chanson 2016); 5 sources unverifiable (Renehan 2004 not_oa; Friedrich 2011 not_oa; van Bunderen 2010, Svensson 2012, Fontana 2008 failed download)

Corrections made:

• Vitale 2012 cohort label: “Ashkenazi longevity families” → “Italian multicenter longevity cohort (Milan, Bologna, Florence)” — this is a different cohort from Milman 2014; the Ashkenazi label was a misattribution. Corrected in body text, footnote, and frontmatter training-cohort field.
• Watts 2022 MR OR values: bare point estimates only → full 95% CIs and p-values added for all three endpoints (overall OR 1.07 [1.00–1.15, P=0.05]; aggressive OR 1.10 [1.01–1.20, P=0.04]; early-onset OR 1.13 [0.98–1.30, P=0.08, non-significant). Early-onset non-significance now explicitly flagged in body and MR section.
• Watts 2022 publication year: cited as “2022” throughout → corrected to “2022/2023” with clarification that the paper was accepted May 2022 but published in IJE 2023 Vol 52(1).
• Watts 2022 colocalization: PP4=99% was correct (SuSiE method) but footnote now clarifies this is SuSiE result; conditional iterative regression gave PP4=0.72.
• Fontana 2016 footnote: Enriched with p-values from Table 4 (between-group P=0.005 for IGFBP-1 at 24 mo; between-group P=0.919 for IGF-1 — confirming null result); protein intake increase in CR group noted as mechanistic explanation.
• Milman 2014 footnote: Enriched with median age of cohort (~96.7–96.8 yr), n per group (93 low, 91 high), cancer subgroup survival details (49.6 vs 20.7 months), and no-malignancy subgroup result (P=0.42, trend only).
• Chanson 2016 footnote: archive status updated; assay names listed; n=911 confirmed; up-to-2.5-fold inter-assay difference noted.
• Renehan 2004, Friedrich 2011, van Bunderen 2010, Svensson 2012, Fontana 2008: tagged no-fulltext-access in footnotes; HR values remain as-extracted pending future verification.
• Fontana 2008 body text: minor note added about discrepancy in starting protein value (1.67 vs 1.6 g/kg/day) vs. Fontana 2016 summary; flagged for future verification.
• Banner: removed (⚠️ auto-extraction banner cleared).

Pages unverifiable (closed-access or download failure):

• Renehan 2004 (Lancet, not_oa) — tagged no-fulltext-access; OR values corroborated by secondary citation in Fontana 2016
• Friedrich 2011 (GH IGF Res, not_oa) — tagged no-fulltext-access
• van Bunderen 2010 (JCEM, failed download) — tagged no-fulltext-access
• Svensson 2012 (JCEM, failed download) — tagged no-fulltext-access
• Fontana 2008 (Aging Cell, failed download) — tagged no-fulltext-access

Downstream pages that may need updates:

• igf-1 — carries Milman 2014 citations; if it repeats the Vitale 2012 Ashkenazi label, correct it
• insulin-igf1 — may reference Watts 2022; ensure early-onset non-significance is noted
• deregulated-nutrient-sensing — may reference Fontana 2016 CR-IGF-1 null result; ensure accuracy
• mk-677 — cross-links confirmed intact

[2026-05-09] verify — biomarkers/chip-clonal-hematopoiesis-biomarker.md

Pages verified: 1

• biomarkers/chip-clonal-hematopoiesis-biomarker.md — 5 primary source PDFs read end-to-end (Jaiswal 2014, Genovese 2014, Bick 2020, Niroula 2021, Vlasschaert 2023, Dorsheimer 2019)

Corrections made:

• Niroula 2021 n: “total not specified” → n=55,383 CHIP WES (46,706 UKB + 8,677 MGBB); M-CHIP HR=7.0 and L-CHIP HR=4.2 for lineage-specific malignancies added to body and footnote
• Niroula 2021 cardiovascular nuance added: L-CHIP NOT associated with CAD or mortality; only M-CHIP (large clones) confers CV risk
• Vlasschaert 2023 methodology clarified: “MR approach” → “genetic proxy (IL6R p.Asp358Ala rs2228145-C)”; non-DNMT3A CHIP specificity of the protective effect added
• Dorsheimer 2019 HR attribution corrected: 2.1× HR is specifically for TET2 or DNMT3A carriers (not all CHIP), as stated in Table 2 multivariable Cox model; KDM6A and BCOR as 3rd/4th most common drivers added
• Dorsheimer 2019 footnote: archive status updated from “pending” to “verified in archive”
• Bick 2020 footnote: archive status updated from “pending” to “verified in archive”
• Sano 2018 body text: no-fulltext-access tag and caveat added since paper is not_oa
• training-cohort frontmatter: Niroula n updated to 55,383

Pages unverifiable (closed-access):

• Sano 2018 (doi:10.1161/CIRCRESAHA.118.313225) — tagged no-fulltext-access; mechanism claim retained with caveat
• Svensson 2022 (doi:10.1001/jamacardio.2022.0386) — previously noted as failed download on phenotypes/clonal-hematopoiesis; unchanged

Downstream pages that may need updates:

• clonal-hematopoiesis — carries the primary quantitative home for Jaiswal/Genovese data; no corrections to those numbers were found so no propagation needed
• tet2 — Niroula cardiovascular nuance (L-CHIP not CV-risk; only M-CHIP) may warrant a note
• senolytics or canakinumab — if they cite Vlasschaert 2023, the non-DNMT3A specificity of IL-6R protection should be noted

[2026-05-09] R36 close — peptide-therapeutics cluster (12 compounds + CAMP parent + class MOC)

User-requested round to seed and verify 12 peptide-class compounds currently under FDA PCAC 503A bulks-list review (2024–2026 cycle). Two parallel waves of 8 + 5 seeders, then matched verifier waves; class MOC at interventions/pharmacological/peptide-therapeutics.md plus parent protein molecules/proteins/camp.md for LL-37 ship as supporting pages. All 13 atomic content pages verified (verified: true) with partial-scope notes for closed-access primaries.

Pages added

• molecules/compounds/bpc-157.md
• molecules/compounds/tb-500.md
• molecules/compounds/ghk-cu.md
• molecules/compounds/epitalon.md
• molecules/compounds/mots-c.md
• molecules/compounds/dsip.md
• molecules/compounds/dihexa.md
• molecules/compounds/mk-677.md
• molecules/compounds/semax.md
• molecules/compounds/melanotan-ii.md
• molecules/compounds/kpv.md
• molecules/compounds/ll-37.md
• molecules/proteins/camp.md (parent protein for LL-37)
• interventions/pharmacological/peptide-therapeutics.md (class MOC)

Mechanism classes added to frameworks/intervention-classes.md

10 new classes registered by seeders during R36:

• tissue-repair-peptide (founding compound: BPC-157)
• actin-regulatory-peptide (founding: TB-500)
• extracellular-matrix-remodeling (founding: GHK-Cu)
• exercise-mimetic / mitohormesis-inducer (founding: MOTS-c)
• sleep-promoting-neuropeptide / hpa-axis-modulation (founding: DSIP)
• growth-factor-receptor-agonism / hgf-mimetic (founding: Dihexa, with retraction caveat)
• growth-hormone-secretagogue / ghrelin-receptor-agonist (founding: MK-677)
• melanocortin-receptor-agonist (founding: Melanotan II)
• antimicrobial-peptide / membrane-permeabilization (founding: LL-37; covers KPV antimicrobial activity too)
• neurotrophin-induction / neuroprotective-peptide (founding: Semax)

Major findings (verifier-derived)

Retracted / absent primary sources:

• Dihexa — c-Met/HGF-mimetic mechanism paper (Benoist 2014, J Pharmacol Exp Ther 10.1124/jpet.114.218735) is RETRACTED. The two iconic marketing claims — “more potent than BDNF” and “outperformed donepezil” — confirmed ABSENT from any surviving primary source after full-PDF reading of McCoy 2013. McCoy 2013’s discussion section explicitly says donepezil comparison is planned future work. The c-Met mechanism itself was attributed in McCoy 2013 to “(Benoist, Kawas, Harding, unpublished data)” — meaning the mechanism was never published in a surviving paper even before the 2014 retraction.
• Semax — user-reported “22% infarct-volume reduction (2005 study)” claim NOT LOCATABLE in PubMed/Crossref/Europe PMC despite exhaustive search. May originate from Russian-language gray literature.
• Melanotan II — user-reported “2021 melanoma incidence review” NOT LOCATABLE. Closest sources are Habbema 2017 case-report compilation (n=4) and Gilhooley 2021 qualitative user-experience study.
• GHK-Cu — iconic “200→80 ng/mL plasma decline (age 20→60)” traced to Pickart 1973 UCSF PhD thesis (unpublished); not in peer-reviewed literature. Both 2015 + 2018 review papers cite the 1973 thesis as the source.

Disambiguations confirmed:

• Epitalon (synthetic AEDG) ≠ Epithalamin (bovine pineal extract). PubMed abstracts of Khavinson/Morozov 2003 (PMID 14523363) and Korkushko 2006 confirm both human cohorts used epithalamin extract, NOT synthetic epitalon. Conflation is a peptide-clinic marketing pattern. Primary epithalamin-vs-epitalon distinction documented prominently on the page.
• TB-500 (LKKTETQ fragment) ≠ full Thymosin β4 (43-aa TMSB4X). Many “TB-500” claims in marketing actually come from full-Tβ4 studies (Philp 2004 et al.) and don’t apply to the fragment. Rahaman 2024 found Ac-LKKTE metabolite is the active species, not parent TB-500.
• Melanotan II (cyclic, never approved) ≠ Melanotan I / afamelanotide / Scenesse (linear, FDA-approved 2019 EPP). Documented prominently.

Seeder-fabricated outcomes caught (R33 + R28 lessons confirmed):

• MK-677: Adunsky 2011 hip-fracture trial terminated early for CHF safety signal missing from seeder draft (verifier added). Barzilai 2012 review’s “40-70% Ames/Snell dwarf” + “16% Igf1r-KO” lifespan percentages attributed to the review but not in its text — corrected to qualitative framing.
• MOTS-c: Zempo 2021 conclusion REVERSED — wiki said variant “enriched in centenarian cohorts in Japan”; paper actually concludes variant “is unlikely to be involved with exceptional longevity” (7.7% vs 7.5% C-allele frequency, n=27,527).
• Semax: Dmitrieva 2010 neurotrophin claim added NT-4 + p75NTR — paper measures only BDNF/NGF/NT-3 and TrkA/TrkB/TrkC. Removed.
• DSIP: two distinct 1981 Schneider-Helmert papers conflated into one footnote; split.
• TB-500: two fabricated author lists caught (Esposito 2012, Yu 1993). Corrected via Crossref.
• MK-677: Svensson 1998 study design misreported as crossover; actually parallel-group. Age range 18-40 → 18-50.
• MK-677: Stage-IV-vs-REM disambiguation — in young adults ~50% increase is stage IV, REM is >20%; in older adults ~50% is specifically REM (effect inverted by age).
• Mots-c: 4 PMID-only refs had wrong DOIs (Woodhead, Yu, Zheng, Yoon) — all corrected via PubMed/Europe PMC; one had wrong CD-1/C57BL/6 mouse-strain assignment, also corrected.

DOI errors caught (R28 lesson — seeder-DOI memory unreliable):

• Singh 2013 (LL-37): wrong DOI resolved to a cyanophage RNA polymerase paper. Corrected to 10.1074/jbc.M112.440883 via PMID lookup.
• Pickart 1980 (GHK-Cu): wrong DOI resolved to Keski-Oja sarcoma growth factors paper. Corrected.
• Zempo 2021 (MOTS-c): typo aging.202593 → aging.202529.
• Zhang 2019 page numbers (LL-37): 232–245 → 121–136.e5.
• PubChem CID for MK-677: seeder used 9912462 → corrected to 178024.

Hallmark linkage convention (R36)

Established: hallmarks: [] empty unless primary-source-grade evidence supports the link. Speculative hallmark linkage by mechanism analogy (the most common seeder error this round) was caught and cleared by verifiers for DSIP. Final hallmark assignments:

• Hallmark-linked: MOTS-c (mitochondrial-dysfunction + deregulated-nutrient-sensing); MK-677 (deregulated-nutrient-sensing + stem-cell-exhaustion); Epitalon (telomere-attrition + cellular-senescence); KPV (chronic-inflammation); BPC-157 (chronic-inflammation, conservative); LL-37 (disabled-adaptive-immunity); CAMP (chronic-inflammation)
• Empty hallmarks: []: TB-500, GHK-Cu, DSIP, Dihexa, Melanotan II, Semax — no defensible primary-source-grade link.

Forward queue surfaced

Implicit-stub follow-ups for future rounds (queued in ROADMAP):

• Parent-protein pages: TMSB4X (TB-500); MOTS-c-peptide protein page (mtDNA-encoded endogenous biology); BDNF protein page (referenced by Semax); FPR2 (LL-37 receptor); GHSR (MK-677 receptor); MC1R/MC3R/MC4R/MC5R (Melanotan II receptors); HGF (Dihexa — lower priority post-retraction); c-Met / MET (Dihexa — lower priority); VEGFR2 + eNOS + growth-hormone-receptor (BPC-157 mechanism proteins); alpha-MSH (KPV/Melanotan II parent peptide).
• Schema gap: clinical-stage controlled vocabulary lacks an abandoned-phase-1 value. Melanotan II workaround used preclinical. Consider adding to CLAUDE.md.
• PCAC bulks-list primary-source confirmation: all FDA process claims sit at #gap/unsourced pending FDA meeting transcript verification.

Cross-page propagation

• LL-37 page updated: Zhang 2019 cell-type precision (THY1hi PDGFRA+ dFB progenitors; TGF-β1; 1-yr-old C57BL/6 male mice; the suppression operates on the differentiation pathway not direct cathelicidin transcription) — propagated from CAMP verifier finding.

Total agents dispatched

13 seeders + 13 verifiers = 26 agent invocations across two waves. Round opened and closed in single calendar day.

---