log/R4.md — Round 4 entries

Sub-file of log — see parent for index.

[2026-05-04] verify | López-Otín 2023 hallmarks paper (PDF acquired via user manual download)

User provided OA URL for the closed-access (per OpenAlex) López-Otín 2023 hallmarks paper; Cloudflare blocked automated curl retrieval, user downloaded manually to sources/pdfs/lopez-otin-2023-hallmarks-expanding-universe.pdf per the wiki sources convention. Sibling .meta.md records provenance.

Verifier subagent (wiki-verifier definition exists at .claude/agents/wiki-verifier.md but isn’t loaded in current session — fell back to general-purpose with the verifier instructions inlined; subagent definition stays for next session) verified the page end-to-end against the 36-page PDF.

Result: all categorizations confirmed correct. No hallmark-stub category: changes needed.

Page-level changes:

verified: false → verified: true; added verified-date: 2026-05-04, verified-by: claude+user
Removed “Cannot verify against full PDF” banner and #gap/no-fulltext-access tag
local-pdf: updated to point at wiki-stored copy
Added explicit 12-hallmark list (vs Figure 1 wheel)
Added explicit Primary/Antagonistic/Integrative categorization table verified against Figure 1 + Integration section (p. 266)
Added Figure 7 “strata of organismal organization” overlay note (orthogonal axis to Primary/Antagonistic/Integrative)
Added connection to “hallmarks of health” framework (López-Otín & Kroemer 2021)
Added Table 1 human-trial roundup: senolytics (D+Q in pulmonary fibrosis, diabetic kidney disease), NAD+ precursors (NMN, NR), urolithin A, canakinumab (CANTOS), Akkermansia muciniphila, elamipretide

Downstream propagation:

frameworks/hallmarks-of-aging.md — “A note on what’s verified” disclaimer rewritten: both 2013 and 2023 papers now PDF-verified; added mention of the new “strata” overlay axis
ROADMAP.md — added urolithin A, canakinumab, elamipretide to compound priority list; added new “Microbiome entities” section with Akkermansia muciniphila; added study-source attributions to NMN/NR
Hallmark stubs unchanged (categorizations were already correct)

FEATURE_REQUESTS item 9 added to a local paper archive: “manual OA URL override” — would have made this a one-command fetch instead of requiring the wiki sources/ workaround.

[2026-05-04] add | ROADMAP.md coverage tracker

Created with planned-vs-actual checklist organized by entity type. Status flags: [ ] not yet seeded; [x] exists with parenthetical state (stub / drafted / verified / verified-partial).

Initial implicit-stub queue derived from grepping all [[wikilinks]] and identifying targets without corresponding .md files, sorted by inbound-link count. Top demand items:

pathways: mtor (11 refs), p53-pathway (7), insulin-igf1 (6), ampk (6), dna-damage-response (5)
processes: autophagy (6), sasp (5), mitophagy (3)
proteins: mdm2 (5), atm (4), bcl-xl (3), cbp-p300 (3)
phenotypes: sarcopenia (5), immunosenescence (4)
compounds: quercetin (3), dasatinib (3) — completes senolytic family alongside fisetin
interventions: caloric-restriction (4)

The roadmap proposes 5 seeding rounds in priority order (Tier 1 implicit stubs → senolytic family → canonical fillers → phenotypes/diseases → hypotheses).

Surrounding updates:

sops/lint-pass.md Step 3 — added the implicit-stub discovery script that powers the roadmap.
index.md — added link to ROADMAP.

Roadmap is descriptive, not gating. Pages can be created without being on the list; the roadmap is a planning aid.

[2026-05-04] add | wiki-verifier subagent

Created project-level subagent at .claude/agents/wiki-verifier.md that implements the verifying-extraction SOP with context isolation. Reasoning: each PDF read pulls 30–60k tokens into context; doing this in the main conversation for every verification quickly fills the window. The subagent reads in its own context and returns a concise corrections summary.

Updates to surrounding documentation:

sops/verifying-extraction.md — added “Use the wiki-verifier subagent” section explaining the invocation pattern and when to use the subagent vs. inline editing.
sops/lint-pass.md — Step 4 now directs the lint operator to invoke wiki-verifier rather than do verification inline.
CLAUDE.md § AI-extracted vs human-verified content — notes the subagent’s existence.
Memory project_aging_wiki.md — records the subagent so future sessions discover it without searching.

Subagent design notes:

Tools: Read, Edit, Bash, WebFetch (for Crossref fallback when a local paper DOI lookup is missing authors/abstract).
Model: sonnet (verification is detailed but bounded; opus reserved for the main conversation).
Hard rules: don’t fabricate values; read methods + results + discussion (not just abstract); don’t propagate downstream (main agent’s job); don’t modify a local paper archive code.
Output contract: ≤300-word summary covering page verified, sources checked, corrections (X→Y format), unverifiable claims, downstream propagation needed, final verified state.

[2026-05-04] verify | first verification pass against PDFs

Schema convention verified: true|false + ⚠️ banner formalized in CLAUDE.md § AI-extracted vs human-verified content. New SOP sops/verifying-extraction.md. Lint pass updated to surface and run verification (Step 4 in sops/lint-pass.md).

Pages verified (3 of 4 study pages flipped to verified: true):

studies/tyner-2002-p53-mutant-aging.md — many corrections:
- Background: corrected to mixed C57BL/6 × 129/Sv (was incorrectly “B6”)
- Cancer incidence: <6% (was “~0%”); WT confirmed >45%
- Median lifespan: 96 vs 118 weeks (23% reduction); was “~20%”
- Maximum lifespan: 136 vs 164 weeks (added)
- p-value: P<0.0001 for survival curve (was “p<0.05”)
- n: 35 (p53+/m) + 56 (p53+/+) for headline comparison; previously “~70”
- Added: pL53 transgenic = internal replication in same paper
- Added: p53-/m crosses confirmed wild-type-p53-dependence of tumor resistance
- Added: m-allele encodes ~24K C-terminal fragment (mechanism)
- Confirmed: paper itself hypothesizes stem-cell-exhaustion as mechanism
studies/yousefzadeh-2018-fisetin-senolytic.md — significant corrections:
- Disambiguated two distinct in vivo experiments (acute gavage vs chronic chow)
- Lifespan study: F1 C57BL/6 × FVB/n hybrids (NOT pure C57BL/6); start at 85 weeks; 500 ppm chow (~60 mg/kg/day); n=8–9/group; Log rank test
- Acute study: 22–24 month C57BL/6, 100 mg/kg gavage × 5 days
- Cell-type specificity (CyTOF): fisetin clears p16+ MSCs/T cells/NK/endothelial; NOT macrophages/dendritic cells — important nuance previously missing
- Added: human ex vivo evidence (n=3 adipose explants, 20 μM × 48h; reduced SA-β-gal + SASP)
- Half-life: 0.09 h rapid / 3.1 h terminal (extremely short; rationalizes “hit-and-run” mechanism)
- Mechanism shifted from incorrect “Bcl-xL/Bcl-2 inhibitor” to paper-supported “PI3K/AKT/mTOR disruption (SCAP)”
- NCT trial references detached from Yousefzadeh footnote (paper doesn’t cite specific NCTs)
studies/lopez-otin-2013-hallmarks-of-aging.md — minor polish, mostly confirmed:
- Primary (4) / Antagonistic (3) / Integrative (2) categorization confirmed against Figure 6
- Added: explicit causal-cascade hypothesis (Primary triggers → Antagonistic responses → Integrative outcomes) is from the paper, not an external interpretation
- Added: Figure 7 intervention table mapping each hallmark to a strategy
- Removed: speculative “>10,000 citations” claim (not from the paper)

Page that cannot be verified (kept verified: false):

studies/lopez-otin-2023-hallmarks-expanding-universe.md — closed access; banner updated to specify the permanent verification blocker. Categorization of the three new hallmarks (disabled macroautophagy, chronic inflammation, dysbiosis) is from secondary sources and explicitly flagged as unverified.

Downstream propagation:

molecules/proteins/p53.md — Tyner 2002-derived numerics corrected (cancer %, lifespan delta, p-value, n, background). Footnote reformatted.
molecules/compounds/fisetin.md — extensive rewrite of mechanism, dose-response table, and PK sections per Yousefzadeh corrections. Added human ex vivo row.
frameworks/hallmarks-of-aging.md — added “A note on what’s verified” disclaimer at top distinguishing verified 2013 categorizations from unverified 2023 ones.

CLI prerequisite documented: archive not on PATH from arbitrary cwds — verification used directly. SOP setup section already documents the symlink/alias workaround.

[2026-05-04] worked-examples + hallmark stubs

Schema shock-test (worked examples):

studies/yousefzadeh-2018-fisetin-senolytic.md (in-vivo, mouse)
studies/lopez-otin-2013-hallmarks-of-aging.md (review)
studies/lopez-otin-2023-hallmarks-expanding-universe.md (review; closed-access — tagged no-fulltext-access)
studies/tyner-2002-p53-mutant-aging.md (in-vivo, mouse, transgenic)
molecules/compounds/fisetin.md (full content: identity, mechanism, PK, dose-response, human trials, classification, gap markers)
molecules/proteins/p53.md (full content: identity, function, antagonistic-pleiotropy aging story, pathway membership, interactors, pharmacology)

Convention added during shock-test:

All AI-extracted study pages and entity pages get a verified: false frontmatter field plus a top-of-page ⚠️ banner noting “auto-extracted, not verified against PDF.” User flips verified: true and removes the banner after PDF review. (This convention is implicit in CLAUDE.md’s evidence discipline section but worth formalizing — defer to user judgment on whether to add to schema.)

Hallmark stubs (12 total) — all categorized:

Primary (5): genomic-instability, telomere-attrition, epigenetic-alterations, loss-of-proteostasis, disabled-macroautophagy
Antagonistic (3): deregulated-nutrient-sensing, mitochondrial-dysfunction, cellular-senescence
Integrative (4): stem-cell-exhaustion, altered-intercellular-communication, chronic-inflammation, dysbiosis

Integration tested:

and confirmed working from.
successfully fetched Yousefzadeh 2018 (gold OA, via PMC mirror) and López-Otín 2013 (bronze OA, via Cell after retry past a 403 — anti-bot retry works).
Tyner 2002 (closed access) was already on disk from prior Anna’s Archive torrent ingestion — surprising bonus.
López-Otín 2023 is closed access and won’t auto-download — handled per the SOP fallback path.

Known issues surfaced:

archive not on PATH from arbitrary cwds — added Setup section to sops/retrieving-papers.md with symlink/alias instructions.
~2–3s startup overhead per invocation (loads 10.8M-row PMC lookup) — added FEATURE_REQUESTS item 8.

Known intentional debt (don’t lint as errors):

Many [[wikilinks]] in the new pages point to pages that don’t yet exist (e.g., [[mtor]], [[mdm2]], [[autophagy]], [[bcl-xl]], [[caloric-restriction]]). These are seed edges for the next round of content; treat as implicit stubs rather than broken-link errors during the next lint pass.

[2026-05-04] update | a local paper archive integration upgraded

FEATURE_REQUESTS 1 and 2 are shipped — wiki SOPs and CLAUDE.md updated to use them.
Updated sops/retrieving-papers.md: replaced fragile archive search --doi table-parsing with (stdout=path, exit-coded) and (JSON).
Updated sops/extracting-evidence.md: documented Crossref fallback for the two metadata gaps the archive currently has — authors (paper_authors not yet populated by ingest) and abstract (not in schema).
FEATURE_REQUESTS 3 (topic tagging) and 4 (queue prioritization) are still pending. Documented workarounds in sops/retrieving-papers.md § “Workarounds for unshipped features”: the wiki’s studies/ folder is the de-facto “aging-relevant” tag set; bulk downloads loop over grep-extracted DOIs.

[2026-05-04] seed | ampk (wiki-seeder subagent)

added: pathways/ampk.md
canonical IDs confirmed: KEGG hsa04152, Reactome R-HSA-380972 (confirmed via WebFetch to Reactome ContentService), WikiPathways WP1471
archive search unavailable (SQL error — similarity(text, smallint) function missing, likely pg_trgm not installed or search code passes wrong type; fell back to per-DOI )
citations verified via (title confirmed against expected): Hardie 2012 (10.1038/nrm3311), Kim 2011 (10.1038/ncb2152), Gwinn 2008 (10.1016/j.molcel.2008.03.003), Martin-Montalvo 2013 (10.1038/ncomms3192)
archive DOI integrity issue observed: many DOIs return mismatched titles (e.g., 10.1073/pnas.0700908104 returns a xylosyltransferase paper; 10.1126/science.1076557 returns an astronomy paper). Only four DOIs with confirmed matching titles were cited.
implicit stubs created: prkaa1, prkaa2, prkab1, prkab2, prkag1, prkag2, lkb1, camkk2, tsc1-tsc2, raptor, ulk1, acc1, pgc1a, energy-stress, fatty-acid-oxidation, mitochondrial-biogenesis, metformin (implicit), caloric-restriction (implicit)
gaps surfaced: needs-human-replication (longevity evidence is invertebrate + mouse pharmacological only), contradictory-evidence (metformin AMPK-dependence in LKB1-null contexts), dose-response-unclear (metformin geroprotective dose), no-mechanism (age-related AMPK decline mechanism in humans), needs-replication (TAK1 as AMPK kinase; Drosophila AMPK lifespan extension)
verifier priority: (1) Hardie 2012 mechanistic claims (subunit structure, nucleotide-binding sites, phosphorylation effects) — highest priority; (2) Gwinn 2008 Raptor phosphorylation sites; (3) Kim 2011 ULK1 phosphorylation sites; (4) Martin-Montalvo 2013 lifespan % and dose

[2026-05-04] seed | p53-pathway (wiki-seeder subagent)

added: pathways/p53-pathway.md
canonical IDs confirmed: KEGG hsa04115, Reactome R-HSA-3700989 / parent R-HSA-5633007, WikiPathways WP1742
archive search unavailable (DB connection BAD — external drive offline); fell back to per-DOI for DOI verification
citations verified via : Tyner 2002 (10.1038/415045a), Haupt 1997 (10.1038/387296a0), Levine & Oren 2009 (10.1038/nrc2723), Rufini 2013 (10.1038/onc.2012.640), Bieging 2014 (10.1038/nrc3711), Maier 2004 (10.1101/gad.1162404), Vousden & Lane 2007 (10.1038/nrm2147), Lane 1992 (10.1038/358015a0)
two DOIs (el-Deiry 1994 and Nakano 2001) returned wrong titles — included as bare DOIs with unsourced caution; verifier should cross-check
gaps surfaced: archive search broken (type-cast bug in ILIKE query when --field used with integer column); implicit stubs created for atm, atr, chek1, chek2, arf, p21, puma, bax, noxa, mdm2, apoptosis-pathway, dna-damage-response, insulin-igf1, oncogenic-stress, ribosomal-stress, cell-cycle-arrest, dna-repair
schema note: apoptosis-pathway used (not apoptosis) per CLAUDE.md convention that process pages get bare names but pathways may use -pathway suffix when name-collision risk exists — consistent with p53-pathway

[2026-05-04] init | scaffold

Created directory skeleton: frameworks/, hallmarks/, pathways/, molecules/{proteins,compounds,metabolites}/, processes/, cell-types/, tissues/, organ-systems/, phenotypes/, interventions/{pharmacological,lifestyle,gene-therapy,stem-cell-therapy}/, model-organisms/, studies/, hypotheses/, gaps/, sources/{images,data,videos,pdfs}/, sops/
Wrote schema: CLAUDE.md (page types, citation rules, gap markers, lint protocol)
Wrote initial MOCs: frameworks/hallmarks-of-aging.md, frameworks/sens-damage-categories.md
Wrote model-organisms/_extrapolation-guide.md (extrapolation rubric)
Wrote 8 SOPs covering paper retrieval, evidence extraction, and pathway/protein/compound/aging-specific database searches
Wrote ~/dev/a local paper archive/FEATURE_REQUESTS.md with 7 prioritized feature requests for the sibling project
Decisions:
- Both Hallmarks of Aging and SENS adopted as co-equal navigational frames
- Atomic-entity pages are the single source of truth; framework pages don’t duplicate facts
- Studies get full pages only when claims are extracted from them (bare DOI footnote otherwise)
- PDFs are NOT duplicated from a local paper archive; sources/ holds only media unavailable there

[2026-05-04] verify | processes/sasp.md

Pages verified: 1 (processes/sasp.md)

Sources checked:

doi:10.1371/journal.pbio.0060301 (Coppé 2008) — PDF verified end-to-end
doi:10.1016/j.cell.2008.03.039 (Kuilman 2008) — PDF verified end-to-end
doi:10.1016/j.cmet.2015.11.011 (Wiley 2016) — PDF verified end-to-end
doi:10.1038/ncb3586 (Glück 2017) — PDF verified end-to-end
doi:10.1016/j.cell.2008.03.038 (Acosta 2008) — partial: PDF not downloadable (OA URL failed); verified via Crossref title/authors and companion-paper text (Kuilman 2008)
doi:10.1016/j.molmed.2010.03.003 (Freund 2010) — partial: PDF not downloadable; verified via PMC full-text web (PMC2879478)
doi:10.1126/science.aaa5612 (Kang 2015) — DOI confirmed via Crossref; PDF not downloadable (green OA, URL returned 0 candidate URLs)

Corrections made:

Coppé 2008 footnote: “human and mouse fibroblasts, epithelial cells” → “5 human fibroblast strains (WI-38, IMR-90, BJ, HCA-2, hBF) + human prostate epithelial cells; 41/120 array proteins altered; SASP requires 4–7 days post-irradiation; in-vivo arm = prostate biopsies”
Kuilman 2008 footnote: “model: human fibroblasts (WI-38, IMR90)” → “human diploid fibroblasts (Tig3(et)/16i HDF); also IMR90 and melanocytes; in-vivo arm: n=20 human colorectal adenoma FFPE samples”
C/EBPβ mechanism: “C/EBPβ activation downstream of IL-6 and IL-8” → corrected to show C/EBPβ is activated by oncogenic stress (upstream), then forms a feedforward loop WITH IL-6/IL-8, not downstream of them; context specified as OIS-specific (BRAF^E600), not general
Paracrine senescence: removed “JAK-STAT activation in recipient cells” as primary Acosta 2008 mechanism; corrected to CXCR2-binding chemokines (IL-8/CXCL8, GRO-family); distinguished IL-6 paracrine role (Kuilman) from chemokine role (Acosta)
cGAS-STING: removed “independently of classical DDR signaling” — Glück 2017 shows cGAS operates in parallel with, not independently of, DDR; clarified CCF source as lamin B1 degradation specifically; removed unsupported “micronuclei or cytoplasmic chromatin bridges” language
MiDAS SASP: “enriched in IFN-related factors and depleted in TGF-β” → corrected to “lacks IL-1-dependent arm (no IL-1β, CXCL1/2, IL-6, IL-8, VEGF); secretes IL-10, TNF-α, CCL27”; clarified AMPK→p53 mechanism more precisely (Ser-15 phosphorylation, NF-κB suppression)
mTOR/rapamycin: removed incorrect attribution to Freund 2010 (that review has no mTOR content); tagged unsourced with candidate primary sources (Laberge/Herranz 2015)
Temporal dynamics: removed unsourced “day 1–3 early phase” claim; now notes Coppé 2008 confirms 4–7 day requirement for robust SASP; early TGF-β phase retained but tagged unsourced (no confirmed primary citation for specific day range)
GATA4: replaced unsourced with confirmed DOI 10.1126/science.aaa5612 (Science 2015, not Cell 2015 as seeder assumed); footnote added with mechanism detail (p62-mediated selective autophagy escape)

Downstream pages potentially needing updates:

molecules/proteins/* pages for IL-6, IL-8, IL-1α (if they cite sasp.md claims about paracrine mechanism)
pathways/cgas-sting.md (if it repeats the “independently of DDR” language)
processes/cellular-senescence.md (if it cites the MiDAS SASP characterization)
interventions/pharmacological/senolytics.md (if it cites the mTOR/rapamycin-SASP claim with Freund 2010)

[2026-05-04] verify | molecules/proteins/atm.md

Pages verified: 1 (partial — two of five sources PDF-verified; three closed-access)

Sources checked:

10.1038/nature01368 (Bakkenist & Kastan 2003) — local PDF read; claims verified
10.1038/ncb1909 (Rodier et al. 2009) — local PDF read; claims corrected
10.1038/nrm3546 (Shiloh & Ziv 2013) — not_oa; Crossref metadata only
10.1126/science.7792600 (Savitsky et al. 1995) — not_oa; Crossref abstract only
10.1126/science.1140321 (Matsuoka et al. 2007) — not_oa; Crossref abstract only

Corrections made:

“inactive homodimer” → “inactive dimer or higher-order multimer” (Bakkenist 2003 abstract: “dimer or higher-order multimer”)
FAT-domain-blocks-kinase mechanism → Ser1981-containing region interacts with kinase domain of opposing monomer (more precise per paper’s biochemical data)
Section header “ATM is required for SASP induction” → “ATM is required for the inflammatory subset of SASP” (Rodier 2009: ATM required for IL-6 and IL-8 subset, not all SASP; 9 of 16 surveyed factors were ATM-independent)
“abrogates SASP cytokine production” → “suppresses a subset of SASP factors — IL-6 ~50-fold, IL-8 ~10-fold” with ATM-independence qualifier
Rodier 2009 footnote cell model: “IMR90 and other” → “HCA2 foreskin fibroblasts (primary); WI-38 (secondary)” — IMR90 is not the primary cell in this paper
Rodier 2009 footnote: enriched with quantitative findings (80–90% ATM depletion, fold-changes, 16-factor survey result)
Matsuoka 2007 footnote: added “>900 regulated phosphorylation sites” (the phosphosite count, distinguished from the >700 protein count in body text)
AT lifespan (~25 y) and heterozygote prevalence (~1%) tagged no-fulltext-access (Shiloh 2013 closed-access)
Three closed-access footnotes updated with not_oa status and no-fulltext-access tags

Domain ranges cross-checked against UniProt Q13315 REST API:

FAT: 1940–2566 ✓
PI3K/PI4K catalytic (kinase): 2686–2998 ✓
FATC: 3024–3056 ✓
HEAT repeats: UniProt does not annotate HEAT repeats for ATM; wiki’s “~1–1939” is a correct inference from FAT domain start; no change needed

Downstream pages potentially needing updates:

processes/cellular-senescence.md — if it uses “ATM is required for SASP induction” language without the subset qualifier
processes/sasp.md — already verified; its ATM claim references Rodier 2009 only in passing (no direct claim to update)

[2026-05-04] verify | phenotypes/sarcopenia.md

Pages verified: 1 (partial — see verified-scope)

Sources checked:

doi:10.1093/ageing/afy169 (Cruz-Jentoft 2019, EWGSOP2) — local PDF downloaded and read end-to-end; diagnostic algorithm and cutpoints verified
doi:10.1093/ageing/afu115 (Cruz-Jentoft 2014, ISI systematic review) — local PDF downloaded and read end-to-end; prevalence range verified
doi:10.1038/nature03260 (Conboy 2005, heterochronic parabiosis) — local PDF read end-to-end; Notch mechanism and framing verified
doi:10.1001/jama.1990.03440220053029 (Fiatarone 1990, nonagenarians resistance training) — not_oa; DOI confirmed via Crossref; n=10, strength +174% ± 31% taken from PubMed abstract only — tagged no-fulltext-access

Corrections made:

typical-onset frontmatter: “Biological onset by age ~40 (1–2%/yr loss)” → “Biological onset by age ~50 (1–2%/yr mass loss after 50)” — EWGSOP2 p.23 specifies “beyond the age of 50 years”
prevalence-65plus frontmatter: “~10–16% (varies by definition and population)” → “1–29% community-dwelling (varies widely by criteria and population); ~10–16% by stricter definitions” — Cruz-Jentoft 2014 reports 1–29% range across 18 studies
Rate-of-decline table row: “~1–2% mass loss/year after 30, ~3% strength loss/year after 60” → “~1–2% muscle mass loss/year and ~1.5–5% strength loss/year after age 50” — matches EWGSOP2 p.23 exact language; original “after 30” age threshold and “~3% strength” figure were not sourced to EWGSOP2
Prevalence prose paragraph: “~10–16%” stated as if canonical → replaced with full 1–29% range per Cruz-Jentoft 2014 Table 1; context note added that the narrower figure reflects stricter European cohort studies
Conboy 2005 mechanism framing: “aged satellite cell defects are largely systemic rather than cell-intrinsic” → corrected to: “aged cells retain intrinsic proliferative capacity, but the old systemic milieu suppresses Notch signalling required for activation” — the paper explicitly states cells retain intrinsic capacity; the wiki’s prior framing overstated the cell-intrinsic-vs-systemic dichotomy
ICD-10 “added 2016” claim removed from body text and footnoted with unsourced — EWGSOP2 confirms ICD-10 coding exists but does not state the year 2016; ICD-11 FB32.Y also unconfirmed in source
Fiatarone attribution: prose “Fiatarone et al. 1990” → proper [^fiatarone-1990] footnote with DOI, journal (JAMA — not NEJM as sometimes cited), n=10, strength gain 174% ± 31%, no-fulltext-access tag
All five footnote references converted to proper schema format with DOI, study design, n, journal, and key findings

New footnotes added:

[^ewgsop2-2019] — EWGSOP2 algorithm and cutpoints (Table 3 values confirmed: grip <27/<16 kg; chair stand >15 s; gait ≤0.8 m/s; SPPB ≤8; TUG ≥20 s)
[^ewgsop2-2019-icd] — ICD recognition claim sourced to EWGSOP2 with unsourced for year and ICD-11 code
[^ewgsop2-2019-rateofloss] — rate-of-decline figures sourced to EWGSOP2 p.23 (acknowledging they cite [129], not primary data)
[^cruz-jentoft-2014-prevalence] — Cruz-Jentoft 2014 ISI systematic review; community 1–29%, long-term care 14–33%
[^conboy-2005] — Conboy 2005 Nature; young/aged C57Bl/6 mice; Notch mechanism
[^fiatarone-1990] — Fiatarone 1990 JAMA; n=10; +174% strength; closed-access

Unverifiable claims:

ICD-10 M62.84 adoption year “2016” — not stated in any source verified; flagged unsourced
ICD-11 FB32.Y — not confirmed in any source; flagged unsourced
Bimagrumab Phase 2 trial details — no primary trial citation added; table entry retained with generic description; needs NCT citation
Falls 1.5–3× risk, mortality 2–3× risk — no primary source footnotes; retained as unsourced for next lint pass
Motor neuron loss “~25% between ages 25 and 75” — no citation; retained as unsourced

Downstream pages to check:

Any page citing sarcopenia prevalence “10–16%” — may need updating to 1–29% range per Cruz-Jentoft 2014
satellite-cells page if it cites Conboy 2005 parabiosis claims (verify framing matches)
stem-cell-exhaustion hallmark page if it references the Conboy 2005 mechanism
senolytics.md if it cites sarcopenia-specific trial evidence

[2026-05-04] verify | pathways/p53-pathway.md

Pages verified: 1 (partial — see verified-scope in frontmatter)

All 10 cited DOIs resolved via DOI lookup + Crossref. 2 DOIs were wrong; all other 8 confirmed correct by title match. PDFs read: Haupt 1997, Vousden/Lane 2007 NRM, Levine/Oren 2009 NRC (all end-to-end), plus Lahav 2004 Nature Genetics for the corrected pulse claim. 5 corrections applied:

el-Deiry DOI corrected: doi:10.1016/0092-8674(94)90273-9 resolves to a Drosophila Jun paper. Corrected to el-Deiry 1993 Cell WAF1 paper (doi:10.1016/0092-8674(93)90500-p); footnote key updated [^eldeiry1994] -> [^eldeiry1993]; in-text reference updated.
Nakano PUMA DOI corrected: doi:10.1016/s1097-2765(01)00373-x failed to resolve. Correct DOI: doi:10.1016/s1097-2765(01)00214-3 (Nakano & Vousden, Mol Cell 7:683-694, 2001).
“~500 target genes” corrected: Levine & Oren 2009 says “hundreds of genes” and “many dozens” — not “~500.” Changed to “hundreds of target genes” + unsourced tag for specific number.
“~5.5h pulse period” corrected and re-attributed: not stated in Levine & Oren 2009. Lahav 2004 (doi:10.1038/ng1293) reports first pulse width 350 +/- 160 min (~5.8 h), inter-pulse interval 440 +/- 100 min (~7.3 h). Text updated with real numbers; new [^lahav2004] footnote added; wrong attribution to [^levineoren2009] removed.
Haupt 1997 half-life caveat: “20-30 min” consensus not explicitly stated in this letter; footnote updated with caveat + cross-pointer to companion Kubbutat et al. 1997 (doi:10.1038/387299a0).

Pages unverifiable (closed-access): doi:10.1038/onc.2012.640 (Rufini 2013, not_oa); doi:10.1038/nrc3711 (Bieging 2014, pending). Qualitative claims retained.

Downstream pages to check: p53 for “~500 target genes” or “~5.5h pulse”; cellular-senescence for OIS/SASP framing from rufini2013.

[2026-05-04] verify | processes/mitophagy.md

Pages verified: 1 (partial — see verified-scope in frontmatter)

PDFs read end-to-end: Narendra et al. 2008 J Cell Biol (doi:10.1083/jcb.200809125), Onishi et al. 2021 EMBO J (doi:10.15252/embj.2020104705), Singh et al. 2022 Cell Rep Med (doi:10.1016/j.xcrm.2022.100633). Lazarou 2015 (not_oa), Fang 2019 (pending), Chan 2012 (pending) not verified.

Corrections applied (6):

Wrong author attribution on review paper: footnote [^ding2021] attributed to “Ding/Bhatt” — actual authors are Onishi, Yamano, Sato, Matsuda & Okamoto 2021 (EMBO J). Footnote key renamed [^ding2021] → [^onishi2021] throughout page; footnote text corrected with proper authorship.
Mechanism table overattribution to Narendra 2008: PINK1 autophosphorylation sites (Ser228/Ser402), phospho-ubiquitin Ser65, and allosteric Parkin activation are post-2008 discoveries not in Narendra 2008. Table restructured to correctly attribute the founding observation (Parkin recruited to depolarized mitochondria) to Narendra 2008 [^narendra2008], and the molecular detail (phospho-ubiquitin mechanism) to [
Narendra 2008 footnote corrected: added explicit note that paper does not characterize PINK1, phospho-ubiquitin, or downstream receptors; added full cell-system list (HeLa, HEK293, MEFs, rat neurons).
FUNDC1 regulation: wiki stated only “CK2 (inhibitory phosphorylation at Ser13)” — omitted Src kinase phosphorylation at Tyr18, also inhibitory under normoxia. Added Src/Tyr18 per Onishi 2021. Citation [^onishi2021] added.
Singh 2022 n=66 error: wiki stated “n=66 middle-aged adults” — paper randomized n=88 (ITT=88, 79 completed). Three arms: placebo n=29, UA 500 mg n=29, UA 1000 mg n=30. Corrected to n=88 randomized/79 completed throughout (interventions table and detail section).
Singh 2022 dose-response conflation: wiki stated “improved muscle strength, VO2 max… at 500 mg/day” — aerobic-endurance and physical-performance (VO2, 6MWT, cycling) improvements were primarily dose-dependent and significant only in the 1000 mg group; the pre-specified primary endpoint (PPO) was not significant vs. placebo. Muscle-strength improvement was significant at both doses. Detail section rewritten with accurate per-group results and p-values.

Pages unverifiable (closed-access or pending): Lazarou 2015 (not_oa — OPTN/NDP52 primary receptor claim unverified); Fang 2019 pending (AD-mitophagy claim unverified); Chan 2012 pending (fission/mitophagy relationship claim unverified).

Downstream pages to check: urolithin-a compound page (may carry n=66 or single-dose framing); urolithin-a if it exists.

[2026-05-04] verify | molecules/compounds/quercetin.md

Page verified: molecules/compounds/quercetin.md

Sources verified: All four PDFs downloaded and read end-to-end.

Zhu 2015 (10.1111/acel.12344) — Aging Cell — verified
Justice 2019 (10.1016/j.ebiom.2018.12.052) — EBioMedicine — verified
Hickson 2019 (10.1016/j.ebiom.2019.08.069) — EBioMedicine — verified
Li 2016 (10.3390/nu8030167) — Nutrients — verified

Corrections made (13):

Wrong serpin attribution (critical): Wiki assigned serpins (SERPINE1/PAI-1) as quercetin’s SCAP targets in HUVECs. Zhu 2015 shows serpins (PAI-2/SERPINB2) are preadipocyte SCAP nodes targeted by dasatinib. Quercetin’s HUVEC targets are EFNB1/BCL-xL and PI3KCD. Corrected in mechanism section and D+Q complementarity table.
Frontmatter mechanisms field: Removed serpin-inhibitor; added bcl-xl-inhibitor to match Zhu 2015.
Justice 2019 quercetin dose: 1,000 mg/day → 1,250 mg/day (250 mg capsules x 5/day per paper Methods).
Justice 2019 primary vs. secondary endpoints: Primary endpoints were retention/completion rates (feasibility). Functional measures (6MWT, gait speed) were secondary. Corrected.
Justice 2019 functional outcomes quantified: 6MWT +21.5 m (p=0.012), gait speed +0.12 m/s (p=0.024), chair-stands -2.2 s (p=0.013), SPPB p=0.003. Null finding added: pulmonary function unchanged.
Hickson 2019 outcomes quantified: p16 -35% (p=0.001), p21 -17% (p=0.009), SA-bgal -62% (p=0.005), macrophages -28% (p=0.0001). SASP factors and adipocyte progenitor findings added.
Mouse group sizes corrected: ~10/group → N=6-9/group (aged C57BL/6); N=8/group (cardiac); N=7-8/group (Ercc1 mice).
Capers quercetin content: ~180 mg/100g → 234 mg/100g (Li 2016 citing USDA database).
Plasma half-life: Removed unsourced fisetin comparison; confirmed Li 2016 states average terminal half-life = 3.5 h (11-28 h range refers to metabolites).
NLRP3 claim deattributed: Li 2016 does not discuss NLRP3. Tagged unsourced.
Citation counts removed from footnotes (volatile, not in source papers).
Footnotes enriched: patient demographics, exact doses, biopsy timing added.
Banner removed; verified flag flipped to true (partial scope: canonical DB fields not re-verified).

Downstream propagation needed:

molecules/compounds/dasatinib.md — may have inherited wrong serpin/quercetin SCAP attribution
processes/apoptosis-pathway.md — cited Zhu 2015; check SCAP target descriptions
Study pages for these four papers (if they exist) — check frontmatter n-subjects against corrected values

[2026-05-04] verify | molecules/proteins/bcl-xl.md

Page verified:

Sources checked:

10.1111/acel.12344 (Zhu 2015) — local PDF, fully read and verified
10.1038/ncomms11190 (Yosef 2016) — downloaded during this pass, fully read and verified
10.18632/aging.101202 (Zhu 2017) — downloaded during this pass, fully read and verified
UniProt Q07817 — REST API verified (domain positions, isoform lengths, PTM enzyme attributions)
10.1016/0092-8674(93)90508-n (Boise 1993) — download failed (bronze OA) no-fulltext-access
10.1016/s1470-2045(10)70261-8 (Wilson 2010) — pending (green OA) no-fulltext-access
10.1038/nm.3048 (Souers 2013) — not_oa no-fulltext-access

Key corrections (10):

Navitoclax misattributed to Zhu 2015: that study tests dasatinib + quercetin only; BCL-xL identified via siRNA, navitoclax not tested. Body text and footnote corrected.
Yosef 2016 cell model: “primary mouse embryonic fibroblasts (MEFs)” → IMR-90 human fibroblasts as primary in vitro model.
Yosef 2016 in vivo model: “aged mice / healthspan improvements” → targeted senescence induction in young mice (lung irradiation + p14ARF transgenic skin). No chronologically aged mice; no healthspan metrics.
IMR90 SCAP node: “BCL-2, BCL-w” (citing Zhu 2015) → BCL-W + BCL-XL dual (synergistic; 53% viability reduction), citing Yosef 2016.
Preadipocyte SCAP node: “BCL-2 (primary); BCL-xL secondary” → EFNB1, EFNB3, PI3KCD, p21, PAI-2 per Zhu 2015 Fig 1D; BCL-xL siRNA not senolytic in preadipocytes.
A1331852 cell-type scope: “senolytic in HUVECs” → senolytic in HUVECs AND IMR90; NOT preadipocytes (Zhu 2017).
Fisetin cell-type scope: “senolytic across multiple cell types” → senolytic in HUVECs only, not IMR90 or preadipocytes (Zhu 2017).
BCL-xS length: “~170 aa” → 188 aa (UniProt Q07817).
Domain residue positions: BH4 (2–25→4–24), BH3 (90–98→86–100), BH1 (136–155→129–148), BH2 (185–196→180–195), TM (212–233→210–226). All per UniProt Q07817.
Caspase cleavage enzyme: “executioner caspases” → “caspase-1 and caspase-3” per UniProt Q07817.

Unverifiable claims:

Boise 1993: isoform discovery and BCL-xL apoptosis mechanism. no-fulltext-access
Wilson 2010: 29/94 grade 3–4 thrombocytopenia; 325 mg/day recommended dose. no-fulltext-access
Souers 2013: venetoclax BCL-2 selectivity and platelet sparing. no-fulltext-access

Downstream propagation needed:

interventions/pharmacological/senolytics.md — SCAP table and navitoclax Zhu 2015 attribution may carry same errors
molecules/compounds/quercetin.md, molecules/compounds/dasatinib.md — BCL-xL SCAP claim should be cell-type-limited (HUVECs only)
processes/apoptosis-pathway.md — Yosef 2016 MEF/aged-mouse framing may be present
molecules/compounds/fisetin.md — multi-cell-type senolytic claim wrong per Zhu 2017 (HUVECs only)

[2026-05-04] verify | molecules/proteins/raptor.md

Page verified: molecules/proteins/raptor.md Sources checked:

10.1016/s0092-8674(02)00808-5 (Kim 2002) — local PDF, fully verified (8 pages)
10.1016/j.molcel.2008.03.003 (Gwinn 2008) — local PDF, fully verified (13 pages)
10.1016/s0960-9822(03)00329-4 (Schalm 2003) — local PDF, fully verified (6 pages)
10.1074/jbc.c200665200 (Nojima 2003) — local PDF, fully verified (4 pages)
10.1016/j.cmet.2008.10.002 (Bentzinger 2008) — local PDF, fully verified (10 pages)
10.1016/j.cmet.2008.09.003 (Polak 2008) — local PDF, fully verified (10 pages)
10.1074/jbc.M113.482067 (Coffman 2014) — local PDF, fully verified (6 pages)
10.1038/s41580-019-0199-y (Liu & Sabatini 2020) — verified via PMC7102936 full text (WebFetch)

Key corrections:

4E-BP1 TOS motif location: “N-terminal” → “C-terminal (FEMDI motif, ~residues 114–118)” per Schalm 2003 Fig 1A and Nojima 2003 Fig 2. S6K1 TOS motif (N-terminal) was correct.
Bentzinger 2008 Cre driver: “MCK-Cre” → “HSA-Cre (human skeletal actin promoter)” — confirmed in paper methods and discussion.
Bentzinger 2008 phenotype: “impaired oxidative capacity, and accumulation of dysfunctional mitochondria” → “loss of oxidative capacity (reduced PGC1α, PPARγ, COX IV), and loss and mislocalization of intermyofibrillar mitochondria.” The paper shows mitochondria are lost/mislocalized, not accumulated.
mTORC1 shape: “rhombus-shaped particle” → “lozenge-shaped particle” per Liu & Sabatini 2020 cryo-EM/crystallographic description.
Rapamycin mechanism: “sterically occludes access by Raptor” → “partially occludes substrate entry into the kinase active site” (rapamycin + PRAS40 together) per Liu & Sabatini 2020.

Unverifiable claims: None — all 8 cited sources were accessible.

Canonical-DB identity fields not re-verified: UniProt Q8N122, NCBI Gene 57521, HGNC 30287, Ensembl ENSG00000141564 — recommend cross-check on next lint pass.

Downstream propagation needed:

pathways/mtor.md — may carry the “rhombus-shaped” mTORC1 architecture claim; check and correct to “lozenge”
molecules/compounds/rapamycin.md — rapamycin mechanism description may use the same “Raptor-occlusion” framing; correct to “partial substrate-entry occlusion”
processes/autophagy.md — ULK1 Ser757 phosphorylation by mTORC1/Raptor may be cited; cross-check phrasing
studies/gwinn-2008-ampk-raptor-checkpoint.md — if this study page exists, confirm Ser722/Ser792 claims are accurate (they are)

[2026-05-04] verify | molecules/proteins/fkbp12.md

Page verified: molecules/proteins/fkbp12.md Sources checked:

10.1074/jbc.270.2.815 (Sabers 1995) — local PDF downloaded (8 pages), fully verified
10.1038/nature08221 (Harrison 2009) — local PDF, fully verified (5 pages)
10.1016/j.cell.2012.03.017 (Laplante & Sabatini 2012) — local PDF downloaded (20 pages), key sections verified
10.1126/science.273.5272.239 (Choi 1996) — closed-access (not_oa); Crossref abstract only
10.1038/369756a0 (Brown 1994) — closed-access (not_oa); Crossref metadata only (no abstract)
10.1126/science.1215135 (Lamming 2012) — green OA but download failed (MIT DSpace 403); unverifiable
10.1126/scitranslmed.3009892 (Mannick 2014) — closed-access (not_oa); Crossref abstract only

Corrections made:

Sabers 1995 model: “rat brain + HEK293 cells” → “rat brain + YAC-1 murine T-lymphoma cells” (HEK293 cells not used; that is Brown/Sabatini 1994, not Sabers 1995)
Harrison 2009 lifespan numbers in body text: “14% females, 9% males by median survival; 13–14% by 90th percentile” → correctly separated mean lifespan (13%/9%) from 90th-percentile/maximal (14%/9%)
Harrison 2009 footnote n: “n=~380/group” → “n=1,960 total; 674 controls, ~317–328/treatment group”
Harrison 2009 footnote effect sizes corrected to match paper’s reported means vs 90th-percentile
Laplante 2012 mTORC2 insensitivity mechanism: removed unsupported claim that rictor “sterically occludes the FRB domain”; paper states FKBP12-rapamycin cannot bind intact mTORC2 but the mechanism is not characterized as FRB occlusion by rictor
Laplante 2012 substrate-access model: softened “sterically occludes access of mTORC1 substrates” claim — review explicitly states mechanism of mTORC1 inhibition by FKBP12-rapamycin “is unknown”; added no-mechanism
Added no-fulltext-access tags to Choi 1996, Brown 1994, Mannick 2014, Lamming 2012 footnotes
Sabers 1995 footnote updated: confirmed full-length mTOR cDNA data (2549 aa, 289 kDa)

Unverifiable claims:

K_d ~0.2 nM (binary FKBP12-rapamycin) and K_d ~12 nM (ternary complex) — attributed to Choi 1996 but closed-access; Crossref abstract confirms crystal structure work but not K_d values
Mannick 2014 effect sizes (~20% vaccine response, infection incidence reduction at 1 year) — closed-access; abstract confirms only general immune improvement finding
Lamming 2012 mechanistic claims (mTORC2 deletion + lifespan, Akt Ser473 specifics) — download failed

Final verified state: verified: true with scope (partial — 3 sources closed-access, 1 download failed)

Downstream propagation needed: pathways/mtor.md, molecules/compounds/rapamycin.md may carry the corrected Harrison 2009 lifespan figures and the mTORC2 insensitivity mechanism claim — main agent should check.

[2026-05-04] verify | molecules/proteins/4ebp1.md

Page verified: molecules/proteins/4ebp1.md Sources checked: 5 of 6 (Mannick 2014 closed-access, tagged no-fulltext-access)

Corrections made:

Zid 2009: “~30% median lifespan extension under DR” → 11% (males) and 22% (females) on rich food; overexpression on DR food shows no extension; source conflated rich-food overexpression gain with DR context
Zid 2009: DR-benefit blunting numbers added: control 35%/42% (m/f) vs null 8%/13% (m/f)
Bjedov 2010: “partially but not completely abolished” → rapamycin extension in 4E-BP-null flies is statistically not significant (p=0.4027, Fig. 5C); the effect is abolished, not partially reduced
Thoreen 2012: footnote cell line “HEK293” → MEFs (4EBP1/2^-/-;p53^-/-); corrected body text and mammalian family section to reflect DKO = Torin1-resistant (not sensitive)
Gingras 2001: phosphorylation table Ser65 entry corrected — Ser65 alone is insufficient to release eIF4E; full hierarchy required (per Figure 6 and Discussion)
Footnotes for Gingras 1999, Gingras 2001, Thoreen 2012, Zid 2009, Bjedov 2010 all updated with verified local-availability status and accurate mechanistic summaries

Unverifiable claims:

Mannick 2014 (n=218 RAD001 RCT, ~20% vaccine improvement): cannot verify — closed-access; tagged no-fulltext-access in footnote

Final verified state: verified: true with scope (Mannick 2014 and UniProt identity fields not re-verified on this pass)

Downstream propagation needed: pathways/mtor.md carries the Bjedov 2010 4E-BP/Atg5 dual-mechanism framing — main agent should update the “partially abolished” language there to match the corrected “statistically abolished” characterization.

[2026-05-04] verify | molecules/proteins/tsc1-tsc2.md

Page verified: molecules/proteins/tsc1-tsc2.md

Sources checked (all six PDFs read in full):

10.1038/ncb839 (Inoki 2002) — local PDF, fully verified
10.1016/j.molcel.2012.06.009 (Dibble 2012) — downloaded, fully verified
10.1101/gad.1110003 (Inoki 2003 rheb) — downloaded, fully verified
10.1073/pnas.202476899 (Tee 2002) — downloaded, fully verified
10.1016/s1097-2765(02)00568-3 (Manning 2002) — downloaded, fully verified
10.1016/s0092-8674(03)00929-2 (Inoki 2003 energy) — downloaded, fully verified

Key corrections:

Stoichiometry claim wrong: “2:2:1 heterotrimer (2 TSC1 : 2 TSC2 : 1 TBC1D7)” is not stated in Dibble 2012. The paper reports a near 1:1 TSC1:TSC2 ratio in TBC1D7 immunoprecipitates by LC-MS/MS (Figure 1F); the overall TBC1D7:core stoichiometry is not precisely determined. Corrected to “heterotrimer with 1:1 TSC1:TSC2 core.”
AMPK sites — species numbering clarified: The wiki body text states Thr1271/Ser1387 (human TSC2 numbering). The original Inoki 2003 energy paper uses rat TSC2 and reports the sites as T1227 and S1345. The human equivalents Thr1271/Ser1387 are confirmed by Dibble 2012 via a pTSC2-S1387 phospho-specific antibody under phenformin (AMPK activator) conditions. Body text language verified correct for human numbering; footnote updated to explicitly note the rat-vs-human site correspondence.
TBC1D7 binding domain claim removed: “Binds TSC1’s coiled-coil region” — Dibble 2012 shows TBC1D7 binds to TSC1 (and can associate in absence of TSC2) but does not identify the specific domain within TSC1 that docks TBC1D7. The coiled-coil specification was unsupported; replaced with accurate description from the paper.
TBC1D7 GAP domain description corrected: Added explicit note that TBC1D7’s TBC domain “lacks two of three conserved motifs required for canonical Rab-GAP activity” (per Dibble 2012 discussion), and that no Rab-GAP activity toward Rheb is established.
Lysosomal localization model tagged: The dynamic TSC2 lysosomal recruitment / cytosolic redistribution model is established in post-2012 literature (Demetriades 2014; Menon 2014) and is not demonstrated by the six cited sources. Tagged unsourced with a note that Dibble 2012 observes ~50% TSC2-LAMP2 co-localization under serum starvation but does not characterize the dynamic redistribution.
All six footnotes substantially enriched with accurate model systems, cell lines, and mechanistic findings from the PDFs.
Auto-extraction banner removed.

Confirmed correct (no change needed):

AKT phosphorylation sites Ser939/Thr1462 (human numbering) — confirmed by both Inoki 2002 and Manning 2002 as the primary sites
Rheb as direct TSC2 GAP substrate — confirmed by Inoki 2003 rheb; TSC2 alone (not TSC1) has GAP activity toward Rheb
TSC1+TSC2 suppress mTOR-mediated S6K1 and 4E-BP1 — confirmed by Tee 2002
14-3-3 sequestration mechanism for phospho-TSC2 — confirmed by Inoki 2002 (phosphomimetic TSC2 less active, disrupts TSC1:TSC2 interaction)
TBC1D7 as third core subunit — confirmed by Dibble 2012
TSC2 protein MW ~200 kDa, TBC1D7 ~34 kDa / 293 aa — confirmed by Dibble 2012

Unverifiable claims (not in any of the six verified sources):

GSK3β sites (Ser1337/Ser1341) — review literature claims; no primary citation verified; unsourced retained
RSK1 site (Ser1798) — same; unsourced retained
TSC2 Rap-GAP domain residue numbers (1531–1758) and critical catalytic residues (Asn1643, Arg1681) — not stated in these six sources; likely from structural/TSC disease literature; unsourced (these are in the TSC2 subunit description bullet, not tagged — flag for next lint pass)

Final verified state: verified: true (full-scope for the six cited primary sources; GSK3β/RSK1/structural-domain numerics remain unsourced)

Downstream propagation needed:

pathways/ampk.md — carries the indirect AMPK→TSC2→mTOR arm; the Thr1271/Ser1387 human numbering should be consistent with the correction note here
molecules/compounds/metformin.md — cites TSC2 Thr1271/Ser1387 in the AMPK arm description; the human numbering is correct but the rat-origin of the data (Inoki 2003) should be noted if metformin.md has a detailed footnote for this
Any page citing Dibble 2012 as evidence for 2:2:1 stoichiometry should be corrected to 1:1 TSC1:TSC2 core

[2026-05-04] verify | molecules/compounds/nmn.md

Page verified: molecules/compounds/nmn.md

Sources checked (all 6 PDFs downloaded and read in full):

10.1126/science.abe9985 (Yoshino 2021) — local PDF, fully verified (14 pages)
10.1016/j.cmet.2011.08.014 (Yoshino 2011) — downloaded and verified (9 pages)
10.1016/j.cmet.2016.09.013 (Mills 2016) — downloaded and verified (13 pages)
10.1038/s42255-018-0009-4 (Grozio 2019) — downloaded and verified (pages 1–10 of 27)
10.1111/acel.12461 (de Picciotto 2016) — downloaded and verified (9 pages)
10.1507/endocrj.ej19-0313 (Irie 2020) — downloaded and verified (8 pages)
10.1210/clinem/dgad027 (Pencina 2023) — downloaded and verified (13 pages)

Key corrections:

Yoshino 2021 (priority source):

NCT “not listed” -> NCT03151239 (confirmed on title page)
Design “crossover” -> “parallel-group” (paper is parallel-group, not crossover)
n split added: 12 placebo / 13 NMN (previously omitted)
Muscle insulin sensitivity effect size added: 25+/-7 0ncrease, p<0.01 (previously vague)
Gene names corrected: “SIRT1, PIK3CA, INSR” -> PDGFRbeta, CD90, CD109, COL1A1, COL5A1, COL6A1 (paper’s actual RNA-seq DEGs; PDGF/muscle remodeling pathway)
Muscle NAD+ corrected: steady-state muscle NAD+ did NOT change; metabolites increased (elevated NAD+ turnover, not pool size)
Mitochondrial oxidative capacity corrected to “did not change” (high-resolution respirometry, paper reports no change)

Pencina 2023:

n corrected: “n=32 active, n=16 placebo” -> n=21 MIB-626, n=9 placebo (total n=30)
Duration corrected: “14 days” -> 28 days
Footnote corrected accordingly

Irie 2020:

Study design corrected: “rct (crossover)” -> “single-arm non-randomized sequential dose-escalation”
Plasma NMN measurement corrected: plasma NMN was NOT successfully quantified (frozen plasma caused degradation); only 2-PY and 4-PY metabolites measured

Yoshino 2011:

Two experiments (HFD 3-6 month mice; aged diabetic males 15-26 months) now distinguished; n corrected to 4-15/group

Mills 2016:

Two doses (100 and 300 mg/kg/day) now documented; dose-dependent effects on eye function, physical activity, and mitochondria noted

de Picciotto 2016:

Mouse age corrected: “26-27 months” -> “26-28 months”
n per group corrected: “~10/group” -> “13-22/group”
Mechanistic detail added (aortic NAD+ not detectable in vivo during chronic supplementation; confirmed threefold increase in vitro)

Grozio 2019:

Tissue specificity clarified: jejunum and ileum primarily; sodium-dependent; IC50 values added; patent COI noted

PK section:

Removed unsupported “plasma NMN peaks within 1-2h in humans” (Irie 2020 failed to measure plasma NMN)

Safety section:

Max human trial duration corrected: “14 weeks” -> “10 weeks”

Unverifiable claims:

Canonical-database identity fields (PubChem CID 14180, InChIKey, ChEMBL) not re-verified against databases
The ~30-50

[2026-05-04] verify | molecules/compounds/nmn.md

Page verified: molecules/compounds/nmn.md

Sources checked (all 6 PDFs downloaded and read in full):

10.1126/science.abe9985 (Yoshino 2021) — local PDF, fully verified (14 pages)
10.1016/j.cmet.2011.08.014 (Yoshino 2011) — downloaded and verified (9 pages)
10.1016/j.cmet.2016.09.013 (Mills 2016) — downloaded and verified (13 pages)
10.1038/s42255-018-0009-4 (Grozio 2019) — downloaded and verified (pages 1-10 of 27)
10.1111/acel.12461 (de Picciotto 2016) — downloaded and verified (9 pages)
10.1507/endocrj.ej19-0313 (Irie 2020) — downloaded and verified (8 pages)
10.1210/clinem/dgad027 (Pencina 2023) — downloaded and verified (13 pages)

Key corrections:

Yoshino 2021 (priority source):

NCT “not listed” -> NCT03151239 (confirmed on title page)
Design “crossover” -> “parallel-group” (paper is parallel-group, not crossover)
n split added: 12 placebo / 13 NMN (previously omitted)
Muscle insulin sensitivity effect size added: 25+/-7% increase, p<0.01 (previously vague “significantly improved”)
Gene names corrected: “SIRT1, PIK3CA, INSR” -> PDGFRbeta, CD90, CD109, COL1A1, COL5A1, COL6A1 (paper’s actual RNA-seq DEGs; PDGF/muscle remodeling pathway, not insulin signaling genes)
Muscle NAD+ corrected: steady-state muscle NAD+ did NOT significantly change; metabolites (N-methyl-nicotinamide, 2-PY, 4-PY) increased indicating elevated NAD+ turnover
Mitochondrial oxidative capacity corrected to “did not change” (high-resolution respirometry, paper explicitly reports no change)

Pencina 2023:

n corrected: “n=32 active, n=16 placebo” -> n=21 MIB-626, n=9 placebo (total n=30 randomized)
Duration corrected: “14 days” -> 28 days
Footnote corrected accordingly

Irie 2020:

Study design corrected: “rct (crossover)” -> “single-arm non-randomized sequential dose-escalation”
Plasma NMN measurement corrected: plasma NMN was NOT successfully quantified (frozen plasma caused degradation); only downstream metabolites (2-PY, 4-PY) were measured

Yoshino 2011:

Two experiments now distinguished (HFD mice 3-6 months; naturally occurring diabetic males 15-26 months); n corrected to 4-15/group (varies by assay)

Mills 2016:

Two doses (100 and 300 mg/kg/day) now documented; dose-dependent effects on activity, eye function, and mitochondria noted; male-only cohort flagged

de Picciotto 2016:

Mouse age corrected: “26-27 months” -> “26-28 months”
n per group corrected: “~10/group” -> “13-22/group”
Mechanistic detail added (aortic NAD+ not detectable in vivo; threefold increase confirmed in vitro at 48h)

Grozio 2019:

Tissue specificity clarified: jejunum and ileum primarily (not generic “small intestine epithelium”); sodium-dependent, not chloride-dependent; IC50 added; patent COI noted

PK section:

Removed unsupported “plasma NMN peaks within 1-2h in humans” (Irie 2020 did not measure plasma NMN successfully)

Safety section:

Max human trial duration corrected: “14 weeks” -> “10 weeks” (longest trial is Yoshino 2021 at 10 weeks)

Unverifiable claims:

Canonical-database identity fields (PubChem CID 14180, InChIKey, ChEMBL) not independently re-verified against databases
The “~30-50% NAD+ tissue decline” figure cited to Yoshino 2011 — precise range for all tissues requires supplemental data

Downstream propagation needed:

studies/yoshino-2021-nmn-insulin-sensitivity.md — if it exists, correct gene names, design type, and NCT
pathways/sirtuin.md — if it cites Yoshino 2021 for insulin gene expression, PDGFRbeta attribution should be noted
Any page with Pencina 2023 n=32/n=16 or 14-day claims should be corrected to n=21/9 and 28 days

Final verified state: verified: true (all six primary-source PDFs verified; canonical-database identity fields not independently re-verified)

[2026-05-04] verify | molecules/compounds/urolithin-a.md

Page verified: molecules/compounds/urolithin-a.md

Sources checked:

10.1016/j.xcrm.2022.100633 (Singh 2022 ATLAS RCT) — local PDF, fully verified (21 pages)
10.1038/nm.4132 (Ryu 2016) — green OA; DOI lookup failed twice (EPFL repository returns HTML, not PDF); unverifiable
10.1038/s42255-019-0073-4 (Andreux 2019) — green OA; DOI lookup failed twice (same EPFL repository issue); unverifiable; design and abstract confirmed via ClinicalTrials.gov NCT02655393 + PubMed efetch
10.1039/d0fo01649g (Garcia 2020) — hybrid OA; still pending download at close of pass; species name confirmed from Crossref/Semantic Scholar title and abstract
PubChem CID 5488186 — confirmed via REST API (InChIKey, SMILES, formula, MW, IUPAC all match)
ChEMBL CHEMBL1836264 — not independently re-verified

Key corrections:

Wrong NCT number for ATLAS trial: NCT03464422 → NCT03464500 (NCT03464422 is an unrelated HIV/stigma behavioral study)
Wrong bacterium species name: Gordonibacter pamelaeae → Gordonibacter urolithinfaciens (Garcia 2020 paper is explicitly about the urolithinfaciens species, not pamelaeae)
Singh 2022 population misdescribed: “healthy middle-aged adults” → “untrained, overweight, middle-aged adults (40–64 yr) with low physical endurance (VO₂max <35 mL/kg/min)” — the trial specifically selected for overweight + low fitness; this is material for interpreting generalizability
Singh 2022 significant muscle-strength findings omitted: The wiki summarized the trial as showing only aerobic-endurance gains; the primary reported positive finding was significant leg muscle strength improvements at both doses (hamstring average peak torque: +12% at 500mg p=0.027, +9.8% at 1000mg p=0.029 vs placebo; knee flexion max torque: +10.6% p=0.017 and +10.5% p=0.022)
Singh 2022 per-arm n added: Breakdown confirmed as placebo n=29 / 500mg n=29 / 1000mg n=30; ITT n=88; per-protocol n=74 (>80% compliance)
Singh 2022 6MWT finding added: 6-min walk distance improved by 33.4m in the 1000mg group (within-group p=0.008; vs-placebo p=0.098 trend)
Singh 2022 safety data added: 102 total AEs across 45 participants; adverse event profile dominated by musculoskeletal/connective tissue (largely from muscle biopsy procedure)
Andreux 2019 population wrong: “healthy middle-aged adults (40–65 yr)” → “healthy sedentary elderly (61–85 yr)” per ClinicalTrials.gov NCT02655393; dose arm description updated to reflect actual trial structure (Part A single ascending dose; Part B 250/500/1000 mg/day × 28 days)
Andreux 2019 acylcarnitine claim narrowed: Wiki claimed only 1000mg/day improved acylcarnitines; abstract states 500mg and 1000mg both modulated plasma acylcarnitines
**Ryu 2016 claims tagged no-fulltext-access: All worm lifespan (~45%), genetic epistasis (dct-1/pink-1), rodent dose (400mg/kg), and in-vitro myotube claims flagged as unverified pending PDF access
Interpretation paragraph updated: Rewritten to acknowledge muscle strength as the primary significant positive finding alongside aerobic endurance, rather than framing endurance gains as the sole positive result

Unverifiable claims:

Ryu 2016: worm lifespan extension (~45%), genetic epistasis (dct-1/pink-1 abrogation), rodent n and dose (400mg/kg/day, 6 weeks), human myotube mitophagy flux — all unverifiable (DOI lookup failed; no-fulltext-access)
Andreux 2019: detailed PK (plasma Tmax, Cmax, conjugate ratios), muscle biopsy method, individual subject data — unverifiable (DOI lookup failed; no-fulltext-access)
Garcia 2020: full enzymatic pathway characterization of which species handles which conversion step — species names confirmed from title/abstract only; full mechanistic detail unverified
ChEMBL CHEMBL1836264 — not re-verified against live ChEMBL

Downstream propagation needed:

processes/mitophagy.md — likely cites Singh 2022 with old population description (“healthy middle-aged adults”) and may be missing the muscle strength finding; should inherit NCT correction
studies/singh-2022-atlas-rct-urolithin-a.md — if it exists, should be checked for the NCT correction (NCT03464500), per-arm n, and muscle strength results
Any other page citing Ryu 2016 for worm lifespan or PINK1/Parkin mechanism via urolithin A should carry the no-fulltext-access qualification

Final verified state: verified: true (partial scope — Singh 2022 fully PDF-verified; Ryu 2016 and Andreux 2019 archive-inaccessible; Garcia 2020 abstract-only; ChEMBL not re-checked)

[2026-05-04] ingest | Li 2023 — dual mTOR/NAD+ gerotherapy (bioRxiv preprint)

Source: https://pmc.ncbi.nlm.nih.gov/articles/PMC9882180/ · doi:10.1101/2023.01.16.523975 · bioRxiv preprint (not peer-reviewed) · fetched via WebFetch; local PDF not yet downloaded (archive status: pending).

Added:

studies/li-2023-dual-mtor-nad-gerotherapy.md — new study page with full frontmatter; preprint banner; all key findings, design, limitations, and extrapolation table

Updated:

pathways/sirtuin.md — Cross-pathway interactions table: expanded [[mtor]] row to include SIRT3-FAO-LPIN1 intersection identified by CRISPRa screen [^li2023preprint]; new footnote added; verified-scope updated
pathways/mtor.md — “Connection to other aging pathways” section: new bullet on NAD+/Sirtuin axis (SIRT3/FAO convergence); new footnote [^li2023preprint]; verified-scope updated

Not updated: molecules/compounds/nmn.md, molecules/compounds/nr.md — paper does not discuss NMN or NR.

Gaps surfaced:

LPIN1 is directionally consistent in the CRISPRa screen but sub-threshold — the mTOR arm of the proposed dual-pathway mechanism is under-supported (#gap/needs-replication)
mTOR-NAD+ cross-talk directionality not established by direct molecular epistasis — inferred via shared FAO regulation only
No human evidence; SOD1-G93A model has limited translatability to general aging (#gap/needs-human-replication)
BIOIO-1001 compound page does not yet exist — could be seeded as a stub under molecules/compounds/bioio-1001.md if the compound gains further traction

Preprint caution: All claims flagged as preliminary (#gap/needs-replication); should be re-evaluated after peer review.

[2026-05-04] verify | molecules/proteins/bak.md

Page verified: molecules/proteins/bak.md Sources checked (all 8 PDFs verified end-to-end):

10.1038/374731a0 (Farrow 1995 Nature) — local PDF, fully verified
10.1038/374733a0 (Chittenden 1995 Nature) — local PDF, fully verified
10.1002/j.1460-2075.1995.tb00246.x (Chittenden 1995 EMBO J) — downloaded and verified in full
10.1016/s1097-2765(00)00136-2 (Lindsten 2000 Mol Cell) — downloaded and verified in full
10.1016/j.molcel.2010.01.025 (Chipuk 2010 Mol Cell review) — local PDF, fully verified
10.1016/j.molcel.2006.10.014 (Moldoveanu 2006 Mol Cell) — downloaded and verified in full
10.1074/jbc.m109939200 (Degenhardt 2002 JBC — previously mislabeled as Suzuki 2002) — downloaded and verified in full
10.1111/acel.12445 (Zhu 2016 Aging Cell) — local PDF, fully verified

Key corrections made:

Discovery section: Farrow 1995 clone is 216 aa; Chittenden 1995 clone is 211 aa (canonical); discrepancy documented; Farrow’s yeast two-hybrid shows BAK does NOT homodimerize with itself
Zinc-binding site residues: His160/His164 → D160 (Asp), H164 (His) per Moldoveanu 2006 Fig 4A–C and crystal coordination geometry (Asp not His at position 160)
Moldoveanu 2006 resolution: corrected to 1.47 Å (SAD dataset in Table 1); paper title rounds to 1.5 Å
PUMA classification: removed from “direct activators” list — Chipuk 2010 Fig 1C and text classify PUMA among sensitizer/derepressor BH3-only proteins, not direct activators; role noted as contested
Lindsten 2000 DKO lethality: “die by postnatal day 1” → “die perinatally within ~48 hours; fewer than 10% survive to adulthood”; neurological defects (deafness, circling) added; strain background specified (mixed 129S3×129X1×C57BL/6)
Critical error corrected: DOI 10.1074/jbc.m109939200 resolves to Degenhardt K, Sundararajan R, Lindsten T, Thompson C, White E (2002) — not “Suzuki M, Youle RJ, Tjandra N” as previously attributed. Wrong authors, wrong journal page numbers, wrong experimental claim. Corrected footnote to Degenhardt et al. 2002; true claim (TNF-α pathway requires both BAX+BAK; either alone is sufficient). Suzuki et al. cytochrome c claim lacks a verified source in this archive.
Zhu 2016 IMR90 triple-knockdown: clarified as BCL-2 + BCL-xL + BCL-w (navitoclax’s target set per Fig 4A); navitoclax NOT senolytic in human preadipocytes (cell-type restriction added)
Chittenden 1995 EMBO J footnote: BH3 functional domain residues 67–94 specified

Unverifiable claims:

Suzuki et al. 2000 (Suzuki M, Youle RJ, Tjandra N) cytochrome c isolated-mitochondria claim — different paper, not in archive; cited as “not locally archived”
UniProt Q16611 canonical identity fields (NCBI Gene 578, HGNC 949, Ensembl ENSG00000030110) — not independently re-verified against databases

Downstream propagation needed:

molecules/proteins/bax.md — may cite Chipuk 2010 and should be checked for the same PUMA activator/sensitizer classification issue
interventions/pharmacological/senolytics.md — already verified; Zhu 2016 IMR90 triple-knockdown claim was previously corrected for bcl-2.md context; confirm bak.md correction is consistent
molecules/compounds/navitoclax.md (if it exists) — should reflect cell-type-restricted senolytic activity (not active in preadipocytes)
Any page citing the “Suzuki 2002” footnote by the old attribution should be updated to Degenhardt 2002

[2026-05-04] verify | molecules/proteins/mcl-1.md

Page verified: molecules/proteins/mcl-1.md

Sources checked:

10.1073/pnas.90.8.3516 (Kozopas 1993, PNAS) — downloaded and read in full (5 pages)
10.1038/nature02067 (Opferman 2003, Nature) — local PDF, read in full (6 pages)
10.1101/gad.215855.113 (Wang 2013, Genes & Dev) — downloaded and read in full (15 pages)
10.1111/acel.12445 (Zhu 2016, Aging Cell) — local PDF, read in full (8 pages)
10.1038/s41467-022-29824-1 (Troiani 2022, Nat Commun) — local PDF, read in full (14 pages)
10.1038/s41392-025-02176-0 (Vogler 2025, Signal Transduct Target Ther) — local PDF, read in full (31 pages)
10.1038/nature19830 (Kotschy 2016, Nature) — not_oa; unverifiable (#gap/no-fulltext-access, already noted in wiki)
10.1038/s41418-020-0564-6 (Shahbandi 2020, Cell Death Differ) — download failed; unverifiable (#gap/no-fulltext-access added)

Key corrections:

MW: “~36 kDa” → “~37 kDa (37.3 kDa predicted; Kozopas 1993)” — Kozopas 1993 explicitly reports 350 aa protein with predicted MW of 37.3 kDa by in vitro translation.
BIM affinity for MCL-1: “~1 nM affinity” → “~74 nM affinity (74 ± 2 nM, BIM-BH3 peptide vs GST-MCL-1 by fluorescence polarization, Opferman 2003)” — 74-fold error; the ~1 nM figure is unsupported by any read source.
Half-life “~30 minutes” (three locations — intro paragraph, Discovery section, PTM section header): the specific “~30 minutes” figure attributed to [^vogler2025] is not stated in Vogler 2025. Vogler 2025 discusses MCL-1 as short-lived and mentions the proteasomal regulation mechanism, but never specifies the time value. All three instances corrected to “often cited as ~20–30 minutes; precise figure not stated in Vogler 2025 unsourced.”
Opferman 2003 lymphocyte phenotype: “rapid and complete loss of lymphocytes” → “profound reduction in both T and B lymphocytes; splenic B cells reduced to ~12% of normal; T cells markedly reduced; residual cells had escaped Cre-mediated deletion.” The paper shows partial reduction with Cre-escape cells surviving; “complete loss” overstates the finding.
Wang 2013 timing: “within weeks of birth” → “all pups died within the first 10 days of birth (constitutive Ckmm-Cre model); within ~3 weeks of tamoxifen in the inducible adult model (Myh-CreER).” Constitutive model kills at ~P10, not “weeks.”
Troiani 2022 footnote model: “human tumor cell lines; scRNA-seq of patient-derived tumor material; mouse xenograft models” → “mouse prostate tumor models (Pten-/- and Pten-/-;Timp1-/- transgenic mice) for scRNA-seq; human prostate cancer cell lines (PC3 shTIMP1, LNCaP) and mouse xenograft models for in vitro/in vivo validation.” No patient-derived material used in scRNA-seq.
Troiani 2022 extrapolation table: “scRNA-seq data is from human tumor material” → “Troiani 2022 scRNA-seq data is from mouse prostate tumor models (Pten-/- mice); human cancer cell lines used for in vitro validation only.”
Shahbandi 2020 footnote: updated from “archive: pending download (bronze OA)” to “archive: download failed — PDF not locally available no-fulltext-access”; quantitative claims from this source noted as unverified.
Kozopas 1993 footnote: updated archive status from “pending download” to “locally downloaded”; added PEST sequence discovery note and protein length/MW detail.
Wang 2013 footnote: archive status updated to “locally downloaded”; updated model description to include both Ckmm-Cre and Myh-CreER; mitochondrial dysfunction confirmed via OCR assays specified; self-verification note removed.

Confirmed accurate items (no correction needed):

BAK as primary MCL-1 substrate — confirmed by Vogler 2025 Fig. 1b and text
NOXA as selective MCL-1 antagonist — confirmed by Opferman 2003 framework (BAD >2,900 nM; NOXA high selectivity) and Vogler 2025 review
BAD negligible affinity for MCL-1 — confirmed in Opferman 2003
Wang 2013 “lethal dilated cardiomyopathy with mitochondrial dysfunction” — confirmed by ECHO (dilated LV, loss of contractility, fibrosis) and OCR data (Table 1); mitochondrial respiratory chain involvement confirmed
Wang 2013 cardiotoxicity mechanistic basis framing — explicitly stated in paper’s discussion
Troiani 2022 MCL-1 dominant anti-apoptotic gene in senescent tumor cells — confirmed; Mcl1 most correlated with senescence score per scRNA-seq
Troiani 2022 S63845 superior to ABT263 for senolytic elimination — confirmed (Fig. 4b, 5)
Zhu 2016 triple siRNA (BCL-2+BCL-xL+BCL-w) required for IMR90 senolysis; MCL-1 not independently dissected — confirmed
Kozopas 1993 institution (Johns Hopkins), journal (PNAS 90:3516–3520), cell line (ML-1), inducer (phorbol ester/TPA) — all confirmed
PEST sequences in MCL-1 N-terminal extension — confirmed by Kozopas 1993 PEST motif analysis

Unverifiable claims:

Any quantitative claims from Kotschy 2016 (S63845 affinity, cancer model results) — not_oa, no local PDF
Any quantitative claims from Shahbandi 2020 (navitoclax + MCL-1i combination data, NOXA biomarker quantification) — download failed
Canonical-database identity fields (UniProt Q07820, NCBI Gene 4170, HGNC 6943, Ensembl ENSG00000143384) — not independently re-verified against live databases

Downstream propagation needed:

molecules/proteins/bak.md (if it exists) — may inherit “BIM ~1 nM affinity for MCL-1” from the same AI extraction; check and correct
pathways/apoptosis-pathway.md — check if it repeats the BIM ~1 nM figure for MCL-1 binding
interventions/pharmacological/senolytics.md — check if it characterizes the Troiani 2022 scRNA-seq model as “patient-derived”; correct to mouse models

Final verified state: verified: true (partial scope — Kotschy 2016 and Shahbandi 2020 unverifiable; canonical-DB identity fields not re-checked)

Final verified state: verified: true (scope: all 8 primary source PDFs verified; canonical-database identity fields not independently re-checked)

[2026-05-04] verify | interventions/pharmacological/senomorphics.md

Page verified: interventions/pharmacological/senomorphics.md

Sources checked:

Herranz 2015 (doi:10.1038/ncb3225) — downloaded and verified against local PDF
Xu 2015 (doi:10.1073/pnas.1515386112) — downloaded and verified against local PDF
CANTOS/Ridker 2017 (doi:10.1056/NEJMoa1707914) — verified against pre-existing local PDF
Laberge 2015 (doi:10.1038/ncb3195) — download failed (no OA URL); no-fulltext-access
Tasdemir 2016 (doi:10.1158/2159-8290.cd-16-0217) — download failed; Crossref abstract only; no-fulltext-access
Salminen 2012 (doi:10.1016/j.cellsig.2011.12.006) — download failed; no-fulltext-access
Lagoumtzi 2021 (doi:10.1016/j.freeradbiomed.2021.05.003) — closed-access (not_oa); no-fulltext-access

Key corrections:

CANTOS fatal infection: “HR 1.31 (p=0.02)” → “0.31 vs 0.18 events/100 person-yr, p=0.02” (NEJM paper reports incidence rates, not HR 1.31)
CANTOS lung cancer HR ~0.67: removed from NEJM-sourced claims; flagged as companion Lancet Oncology paper data; replaced with correct NEJM fatal-cancer data (incidence rate 0.45 vs 0.64/100py, p<0.001 trend)
CANTOS 300 mg threshold: clarified DID NOT meet prespecified threshold (p=0.031 vs required p<0.01058)
Xu 2015 JAK inhibitors: “ruxolitinib and tofacitinib” → tofacitinib not in Xu 2015; corrected to JAK inhibitor 1 (pan-JAK), ruxolitinib (INCB18424, JAK1/2), momelotinib (CYT387)
Xu 2015 in-vivo model: “naturally-aged mice” → “24-month-old C57BL/6 male mice (n=8–9/group)”; dose added (60 mg/kg/day oral); physical function endpoints enumerated
Herranz 2015 mechanism: added ZFP36L1 as key intermediate (MAPKAPK2 phosphorylates ZFP36L1 → inactivates its mRNA destabilizing function → SASP mRNAs stabilized via ARE-mediated stabilization); MAPKAPK2 de novo synthesis reduced ~80% by mTOR inhibition
Herranz 2015 cell model: “human fibroblasts” → IMR90 ER:RAS (OIS), BJ, HFFF2, plus 22-month mouse liver in-vivo validation
Tasdemir 2016: added dual-function nuance — BRD4/SASP required for tumor-suppressive immune surveillance; BET inhibition has dual cost in cancer context (from Crossref abstract; full PDF unverifiable)
All unverifiable footnotes tagged no-fulltext-access; Herranz 2015 and Xu 2015 footnotes updated with accurate detail

Downstream propagation needed:

Any page citing CANTOS with “HR 1.31” for fatal infection or “HR ~0.67 lung cancer” attributed to the NEJM paper — correct to incidence rates / companion Lancet Oncol citation
processes/sasp.md — verify Xu 2015 JAK inhibitor identity and Herranz 2015 ZFP36L1 mechanism
molecules/compounds/rapamycin.md — verify Herranz 2015 ZFP36L1 mechanism is correctly described

Final verified state: verified: true (partial scope — Laberge 2015, Tasdemir 2016, Salminen 2012, Lagoumtzi 2021 unverifiable; claims from these sources tagged no-fulltext-access)

[2026-05-04] verify | model-organisms/drosophila-melanogaster.md

Page verified: model-organisms/drosophila-melanogaster.md Sources checked:

Bjedov 2010 (10.1016/j.cmet.2009.11.010) — local PDF, full read. Verified.
Zid 2009 (10.1016/j.cell.2009.07.034) — local PDF, full read. Verified.
Kapahi 2004 (10.1016/j.cub.2004.03.059) — downloaded during this pass, full read. Verified.
Rogina 2004 (10.1073/pnas.0404184101) — downloaded during this pass, full read. Verified.
Clancy 2001 (10.1126/science.1057991) — closed-access (not_oa); Crossref abstract only.

Corrections made:

Kapahi 2004 mechanism description: “heterozygous for dTOR or with reduced dS6K show ~25% lifespan extension” → corrected to transgenic overexpression of dTSC2/dTOR^FRB^/dS6K^KQ^ via da-GAL4 driver with specific % values (dTSC2: 14%/12% M/F; dTOR^FRB^: 24%; dS6K^KQ^: 22%). The previous description implied heterozygous loss-of-function which was wrong.
Kapahi 2004 footnote n: “~100–200” → “~92–182” per figure legends.
Rogina 2004 driver: “driven from a heat-shock promoter” → corrected to tubulin-GAL4 and ELAV-GAL4 drivers. No heat-shock promoter was used in this paper.
Rogina 2004 effect size: “~59%” → “up to ~57% in females, ~32% in males” (tubulin-GAL4; Table 1 values).
Rogina 2004 footnote n: “~120/group” → “≥149–232/group” per Table 1.
Failure modes section: clarified that Rogina 2004 did NOT use heat-shock promoters; note applies to other papers.
Clancy 2001 chico: homozygote % (up to 48%) verified from Crossref abstract; sex-specific heterozygote breakdown (females ~36%, males ~13%) tagged no-fulltext-access as it cannot be confirmed without full PDF.

Pages needing downstream update:

studies/rogina-2004-sir2-drosophila-cr — if this study page exists, update driver type and effect sizes
molecules/compounds/rapamycin.md — cross-check Kapahi 2004 framing if cited there
hallmarks/cellular-senescence.md or sirtuin-related pages — if Rogina 2004 cited with heat-shock framing

Final verified state: verified: true (partial scope — Clancy 2001 sex-specific heterozygote claim not independently confirmable; tagged no-fulltext-access)

[2026-05-04] verify | model-organisms/caenorhabditis-elegans.md

Page verified: model-organisms/caenorhabditis-elegans.md

Sources checked:

10.1038/35065638 (Tissenbaum & Guarente 2001) — verified against local PDF
10.1038/nature01298 (Holzenberger 2003) — verified against local PDF (pre-verified for sirtuin; cross-confirmed)
10.1073/pnas.95.22.13091 (Lakowski & Hekimi 1998) — downloaded and verified
10.1101/gad.1255404 (Apfeld et al. 2004) — downloaded and verified
10.1073/pnas.0809594106 (Flachsbart et al. 2009) — downloaded and verified
10.1038/366461a0 (Kenyon 1993) — not_oa, closed-access; unverifiable; tagged no-fulltext-access
10.1038/426620a (Vellai 2003) — download failed (bronze OA URL inaccessible); tagged no-fulltext-access
10.1126/science.1087782 (Melendez 2003) — not_oa, closed-access; abstract verified via Crossref; tagged no-fulltext-access

Corrections made:

Holzenberger 2003 males lifespan: “~16% longer” → “~15.9% longer” (exact paper figure)
FOXO3A claim: “GG homozygotes show association” → “minor allele carriers; OR ~1.42 for top SNP rs3800231 in centenarians” (precise per paper)
Flachsbart 2009 footnote n=: “1,762 centenarians/nonagenarians” → “1,031 LLI + 731 controls” (1,762 was total sample including controls, not just LLI)
Apfeld 2004 epistasis claim: “AAK-2 acts upstream of DAF-16” → corrected to “aak-2 and daf-16 act in parallel” (paper explicitly shows double mutant 15% shorter than either single mutant; parallel relationship)
Apfeld 2004 quantitative figures added: aak-2 loss → 12% shorter lifespan; aak-2 overexpression → 13% longer (both from paper Table/Fig 2B)
Lakowski 1998 quantitative figures added: eat-2(ad1116) +57% lifespan over N2; average for ad465 +47% over 7 trials; DR/IIS mechanistic distinctness (daf-16-independent; additive with daf-2) now explicit in page body
Tissenbaum 2001 footnote: fabricated n range “n=80-451/strain” removed; actual per-strain n values from paper substituted
Melendez 2003 autophagy scope: clarified that autophagy requirement was specifically demonstrated for IIS pathway mutants; overgeneralization to “multiple longevity mutations” qualified
Melendez 2003 gene list: added orthologs of APG1, APG7, APG8, AUT10 (named in paper abstract)

Pages unverifiable (closed-access or download failed):

Kenyon 1993 — tagged no-fulltext-access
Vellai 2003 — tagged no-fulltext-access
Melendez 2003 — tagged no-fulltext-access (abstract-level claims cross-checked via Crossref)

Final verified state: verified: true (partial — three sources unverifiable; see verified-scope)

Downstream pages that may need updates:

molecules/proteins/ pages for daf-16, daf-2, aak-2 (if they exist) — AAK-2/DAF-16 parallel relationship correction may propagate
pathways/ampk.md — if it inherits the “upstream of DAF-16” framing from this page
molecules/compounds/ sirtuin-related pages — Tissenbaum n values now corrected
interventions/lifestyle/caloric-restriction.md — eat-2 quantitative lifespan figures updated

[2026-05-04] verify | cell-types/cardiomyocytes.md

Page verified: cell-types/cardiomyocytes.md

Sources checked:

Bergmann 2009 (10.1126/science.1164680) — PDF not downloadable (archive failed, no OA PDF candidate URLs found after 2 retry attempts; status: green OA but Science CDN not indexed). Turnover rate claims verified against Crossref abstract, which states figures verbatim.
Bergmann 2015 (10.1016/j.cell.2015.05.026) — PDF downloaded and read in full.
Anderson 2019 (10.15252/embj.2018100492) — PDF downloaded (PMC) and read in full.
Lewis-McDougall 2019 (10.1111/acel.12931) — PDF downloaded (PMC) and read in full.
Wang 2013 (10.1101/gad.215855.113) — local PDF read in full; consistent with pre-verified mcl-1 page.

Corrections made:

Ploidy claim (CORRECTION, Bergmann 2015): body table said “human adult CMs: ~25–33% mononucleated.” Bergmann 2015 (Fig 1C) reports 73.6%±7.0% mononucleated, 25.5%±7.4% binucleated. The wiki had inverted the mononucleated/binucleated proportions. Corrected to “~73.6% mononucleated, ~25.5% binucleated, ~1.0% trinucleated.”
Turnover rate (enrichment, Bergmann 2015): Added Bergmann 2015’s refined number (<0.3%/yr at age 75) alongside the Bergmann 2009 figure (0.45%/yr). Also added that 2015 confirms cardiomyocyte number is established perinatally and constant thereafter, and that ~39% of LV CMs are exchanged postnatally over a lifetime.
SASP composition (CORRECTION, Anderson 2019): Body text said cardiomyocyte SASP includes “TGF-β, GDF-15, and inflammatory cytokines.” Anderson 2019 identifies a non-canonical SASP in purified aged CMs with Tgfb2, Gdf15, and Edn3 — explicitly noting that canonical inflammatory components (IL-6, IL-1β, Cxcl1) were NOT upregulated in purified CM vs stromal fractions. Corrected in body, niche section, and footnote.
Lewis-McDougall attribution (CORRECTION): Wiki stated “aged mouse hearts contained more senescent cells than young, and co-injection of aged-heart-derived senescent cells with cardiac progenitor cells inhibited progenitor engraftment.” Lewis-McDougall 2019 is a study of cardiac progenitor cells (CPCs) from human hearts, not a whole-heart senescent burden comparison between young and old mice. The injection experiment compares senescent CPCs vs cycling CPCs injected separately (not co-injection). Corrected to accurately describe the CPC-focused human CPC data, NSG mouse MI model, and in vivo D+Q/AP20187 senolytic experiments.
Navitoclax in vivo attribution: Clarified that in vivo navitoclax cardiac senolytic data is from Anderson 2019; Lewis-McDougall uses navitoclax in vitro only, with D+Q and AP20187 for in vivo experiments.
NF-κB SASP claim: Qualified — Anderson 2019 finding that classical NF-κB targets not upregulated in purified aged CMs is inconsistent with a broad “drives inflammatory SASP” claim.
Bergmann 2015 footnote: Updated “n not extracted” to actual sample sizes (n=51 ¹⁴C dating, n=29 stereology) and added quantitative turnover and ploidy data.
Anderson 2019 footnote: Enriched with experimental detail, n per group, SASP factor identities, navitoclax dose, and mitochondrial ROS mechanism.
Lewis-McDougall 2019 footnote: Corrected from “co-injection experiments” to accurate description of the CPC injection paradigm, SASP factors (PAI-1, IL-8, IL-6, IGFBP-3, GM-CSF, CCL-2), EdU/Ki67 quantification post-clearance, and noting navitoclax was in vitro only.

Unverifiable claims:

Bergmann 2009: n=24 individuals (stated in wiki footnote) — cannot confirm without full PDF. All three headline claims (turnover ~1%/yr at 25, ~0.45%/yr at 75, <50% lifetime replacement) match Crossref abstract verbatim; flagged with no-fulltext-access on footnote.
“Decreased Complex I, III, and IV activity” in aging cardiomyocytes — Wang 2013 shows Complex I and IV impairment in MCL-1 KO but this is KO-driven, not aging per se. The claim is under aging features and is not Wang-sourced; it remains unsourced.

Wang 2013 claims confirmed against PDF:

Constitutive Ckmm-Cre model: all pups (n=22) died within 10 days of birth. Correct.
Inducible Myh-CreER model: all mice succumbed within ~3 weeks of tamoxifen, heart failure with dilated cardiomyopathy and loss of myofibrils. Correct.
Mitochondrial dysfunction (Complex I, IV impaired; respiratory control ratio decreased; mtDNA loss) confirmed in KO. Correct.
Consistent with pre-verified mcl-1 page; no conflicts.

Downstream propagation needed:

molecules/proteins/mcl-1.md — verified; no conflicts found with cardiomyocytes.md Wang 2013 claims.
hallmarks/cellular-senescence.md — if it cites Anderson 2019 for CM SASP composition, may need SASP correction (Tgfb2/Gdf15/Edn3, not inflammatory cytokines).
processes/cellular-senescence.md (if exists) — same SASP correction may apply.
Any page citing “cardiomyocytes ~25–33% mononucleated” should be corrected to ~73.6% mononucleated.

[2026-05-04] verify | cell-types/microglia.md

Page verified: cell-types/microglia.md

Sources checked:

Ginhoux 2010 (10.1126/science.1194637) — green OA PDF downloaded and verified end-to-end
Guerreiro 2013 (10.1056/NEJMoa1211851) — bronze OA PDF downloaded and verified end-to-end
Elmore 2014 (10.1016/j.neuron.2014.02.040) — bronze OA PDF downloaded and verified end-to-end
Hammond 2019 (10.1016/j.immuni.2018.11.004) — bronze OA PDF downloaded and verified end-to-end
Keren-Shaul 2017 (10.1016/j.cell.2017.05.018) — bronze OA; Cell.com 403 blocked download; no-fulltext-access on footnote
Jonsson 2013 (10.1056/NEJMoa1211103) — green OA; download stalled (stuck in “downloading” state); OR confirmed via Europe PMC abstract; full methods/results not read
Norden 2013 (10.1111/j.1365-2990.2012.01306.x) — green OA; archive returned “0 candidate URLs after filtering”; download failed; no-fulltext-access on footnote

Corrections made:

Guerreiro 2013 footnote DOI (CRITICAL CORRECTION): wiki footnote cited doi:10.1056/nejmc1306509 — this is an NEJM Correspondence letter, not the original article. Correct DOI is 10.1056/NEJMoa1211851. Corrected throughout.
Guerreiro 2013 footnote authorship: changed “Guerreiro R and Hardy J” to “Guerreiro R et al.” (multi-author paper). Added full sample sizes: 1092 AD cases + 1107 controls (discovery); 1887 cases + 4061 controls (replication). Added precise ORs: R47H OR 4.5 (95% CI 1.7–11.9); all-variant OR 4.59 (95% CI 2.49–8.46, P=9×10⁻⁹).
Guerreiro 2013 cited_by: corrected from ~86 to ~2949 (the nejmc1306509 correspondence had ~86 citations; the NEJMoa1211851 paper has ~2949).
Jonsson 2013 footnote OR (CORRECTION): changed “OR ~2.9” to “OR 2.92 (95% CI 2.09–4.09) discovery; OR 2.90 (95% CI 2.16–3.91, P=2.1×10⁻¹²) combined.” Added full replication cohort description (US, Norway, Netherlands, Germany).
TREM2 section body (CORRECTION): removed vague “~2–4× increased risk” phrasing; replaced with precise ORs from both papers: Jonsson OR 2.90 (95% CI 2.16–3.91), Guerreiro OR 4.5 (95% CI 1.7–11.9). Added context that both papers were published simultaneously in NEJM Vol. 368:2.
Elmore 2014 CSF1R inhibitor identification (CORRECTION): wiki listed “PLX5622, PLX3397” as co-equal in vivo compounds. Elmore 2014 used PLX3397 (290 mg/kg chow) as the primary in vivo compound throughout all chronic studies; PLX5622 is mentioned only in early in vitro dose-response. Reordered to “PLX3397, PLX5622” and added “PLX3397 at 290 mg/kg chow” as the standard protocol.
Elmore 2014 depletion timeline (CORRECTION): wiki said “within 1–2 weeks” — paper shows ~50% reduction by day 3, essentially complete by day 21 (~99%). Corrected to “~3 weeks.”
Elmore 2014 repopulation timing (CORRECTION): wiki said “within ~1 week” for full repopulation. Paper shows microglia continue declining 1–2 days after drug withdrawal, then begin repopulating; microglia numbers equal controls by 7 days. More precisely, the critical repopulation window is 48–72 hr, with full numerical recovery by ~7 days. Corrected to “begins within 48–72 hours; full recovery ~7 days.”
Elmore 2014 footnote model description (CORRECTION): removed “PLX5622/PLX3397 diet” — implies equivalence. Corrected to “PLX3397 (290 mg/kg chow) as primary in vivo CSF1R inhibitor.”
Elmore 2014 repopulation benefit claim (CORRECTION): wiki attributed cognitive improvement in aged mice to “Elmore 2014; subsequent work by Green lab.” Elmore 2014 tested young adult mice (2–18 months) and found NO cognitive deficits from depletion and mildly improved Barnes maze. Aged-mouse repopulation benefit claims are not from Elmore 2014. Corrected and tagged unsourced.
Ginhoux 2010 footnote model system (CORRECTION): footnote said “model: E8.5 lineage tracing, Csf1r-Cre × Rosa26-YFP.” The paper used Runx1-MERCre-MER × Rosa26^R26R-eYFP (tamoxifen-inducible). Csf1r-Cre × Rosa26-YFP was not used. Also corrected temporal claim: progenitors arise before E8.0 (abstract says “before embryonic day 8”); brain seeding occurs E8.5–E9.5 via blood vessels.
Hammond 2019 cell count (CORRECTION): scRNA-seq table header and footnote implied cell counts matched Keren-Shaul’s ~4000. Hammond 2019 profiled 76,149 cells from 41 mice. Added to footnote.
Hammond 2019 aging cluster markers (MAJOR CORRECTION): wiki table listed “Aged-specific | Cd52+, Cd63+” — this is incorrect. Hammond 2019’s aging-enriched clusters are OA2 (Ccl4+, Il1b+, Lpl+, Cst7+; 2–4× increased at P540) and OA3 (Ifitm3+, Irf7+, Rtp4+; interferon-response; small subset). Cd52 appears in the injury-response cluster IR2, not specifically in aging. Cd63 is not identified as an aging marker in Hammond 2019. Table entirely rewritten.
Hammond 2019 number of states (CORRECTION): wiki did not specify a number. Hammond 2019 identified 9 transcriptionally distinct states. Added to footnote and body text.
DAM two-step terminology (CORRECTION): wiki used “early DAM” and “late DAM.” Keren-Shaul 2017’s terminology (as confirmed by Hammond 2019 discussion referencing this paper) is “Stage 1” (TREM2-independent) and “Stage 2” (TREM2-dependent). Updated to use Stage 1/Stage 2 terminology. Added gap tag since Keren-Shaul PDF unverified.
Norden 2013 quantitative claim tagged (CORRECTION): “~2–3× more IL-1β” claim tagged no-fulltext-access — quantification is from this review’s synthesis of primary sources; full text not accessible to verify.
Norden 2013 footnote year: clarified “published 2012/2013” (metadata records 2012; journal vol indicates Neuropathology and Applied Neurobiology).

Unverifiable claims (failed downloads):

Keren-Shaul 2017: DAM Stage 1/Stage 2 marker list, n-cells, precise 5XFAD experimental design — no-fulltext-access on footnote.
Jonsson 2013: OR confirmed via abstract only; methods, sample sizes, exact discovery cohort characteristics — not read from full text; OR values confirmed numerically against Europe PMC abstract.
Norden 2013: all quantitative aging-priming claims — no-fulltext-access.

Downstream propagation needed:

studies/guerreiro-2013-trem2-ad.md — if this study page exists, the DOI in its frontmatter may be 10.1056/nejmc1306509 (wrong) and must be corrected to 10.1056/NEJMoa1211851.
molecules/proteins/trem2.md — if it cites TREM2 R47H OR from Guerreiro 2013, the OR should be 4.5 (95% CI 1.7–11.9) not a vague “2–4×.”
hallmarks/chronic-inflammation.md — if it cites Hammond 2019 aging clusters, Cd52+/Cd63+ markers need correction to OA2/OA3 cluster signatures.
alzheimers-disease.md — if it cites TREM2 R47H OR from Jonsson or Guerreiro, verify precision matches corrected values above.

[2026-05-04] round-4 summary

Round 4 complete: 4a (model organisms tier 1) + 4b (model organisms tier 2) + 4c (extreme-lifespan organisms) + 4d (cell-types tier 1).

Drafted (8 new pages):

4a/b: mus-musculus, rattus-norvegicus, caenorhabditis-elegans, drosophila-melanogaster, saccharomyces-cerevisiae, homo-sapiens
4c: nothobranchius-furzeri, heterocephalus-glaber
4d: hematopoietic-stem-cells, satellite-cells, microglia, cardiomyocytes (first cell-type pages — schema addition cell-ontology-id)

Verified (12 of 12; all to verified-partial or higher):

mus-musculus → verified-FULL (7 PDFs end-to-end)
rattus-norvegicus → verified-partial
caenorhabditis-elegans → verified-partial
drosophila-melanogaster → verified-partial
saccharomyces-cerevisiae → verified-partial
homo-sapiens → verified-partial
nothobranchius-furzeri → verified-partial
heterocephalus-glaber → verified-partial
hematopoietic-stem-cells → verified-partial
satellite-cells → verified-partial
microglia → verified-partial
cardiomyocytes → verified-partial

Most consequential corrections this round:

Killifish GRZ lifespan ~2× inflated: “4 mo” → ~9 weeks median (~3 mo max) per Terzibasi 2009 + Valenzano 2006. Propagated to model-organisms/README.md and _extrapolation-guide.md.
Cardiomyocyte ploidy INVERTED: human adults ~73.6% mononucleated (NOT ~25-33%) per Bergmann 2015 Fig 1C.
Cardiomyocyte SASP non-canonical: Anderson 2019 explicitly distinguishes — Tgfb2/Gdf15/Edn3 only, NO IL-6/IL-1β/Cxcl1 in purified aged CMs. Wiki had attributed canonical inflammatory SASP.
Microglia: wrong DOI for Guerreiro 2013 — nejmc1306509 (Correspondence, ~86 cites) → NEJMoa1211851 (original article, ~2,949 cites). TREM2 R47H ORs 4.5 (Guerreiro) and 2.92 (Jonsson) added precise.
Microglia Hammond 2019 aging clusters wrong: “Cd52+/Cd63+” → OA2 (Ccl4+/Il1b+/Lpl+/Cst7+) and OA3 (Ifitm3+/Irf7+/Rtp4+) — Cd52 is injury-cluster, Cd63 absent from aging signature.
Microglia Elmore 2014 primary compound PLX3397 (NOT PLX5622); depletion ~3 weeks (NOT 1-2 wks); aged-cognitive-benefit was NOT in Elmore 2014.
HSC: Sudo 2000 expansion ~17× CD34⁻KSL gate (NOT ~10× SLAM — Kiel SLAM markers postdate Sudo); Beerman 2010 only 2 subsets (not 3); Jaiswal 2014 prevalence “>70 yr” (NOT >65); CHIP HRs split (1.4 all-cause / 2.0 CHD / 2.6 stroke) not composite “~40%”; Genovese 2014 WES (not WGS).
Satellite cells Conboy 2005 strain: young = C57Bl/Ka-Ly5.2 (NOT C57Bl/6); only old = C57Bl/6. Propagated to sarcopenia.md line 171.
NMR Takasugi 2021 reframed (MAJOR): paper DISPUTES Tian 2013’s 6-12 MDa HMW-HA figure (NMR HA max ~2.5 MDa via SEC/HABP; methodological critique of Alcian Blue). Added contradictory-evidence.
NMR Ruby 2018 caste disaggregation present (wiki had said it wasn’t); breeders ~5-10× lower hazard.
NMR Seluanov 2009 contact-inhibition mechanism direction CORRECTED: p16 mediates EARLY low-density CI (novel layer); p27 mediates LATER regular CI.
Yeast: SIR2-independence requires sir2Δ fob1Δ DOUBLE mutant (Kaeberlein 2004 Fig 2A); sir2Δ alone abolishes CR effect.
Rat: SHR ambiguity RESOLVED — SHR is mice not rats per Anisimov 2010 Table 1 (matches metformin verifier framing); Masoro 1982 reframed from “review” to primary data paper (n=115/group F344, 60% CR, 41% extension).
Mouse: Strong 2016 acarbose ♂“+22%” was wrong attribution (that’s Harrison 2014; Strong 2016 actual ♂+6%); Harrison 2009 dose 600 ppm → 14 mg/kg food (consistent across rapamycin/fkbp12/mtor verifiers); ortholog 85% → 80%.

Schema additions this round:

cell-ontology-id (cell-type frontmatter)
lifespan-ratio-to-human + itp-validated + genome-size-mb + adult-cell-count (model-organism)

Persistent verifier-DB issues:

Cell.com PDFs blocked by Cloudflare 403 (Keren-Shaul 2017, López-Otín 2023, Hammond not affected)
Cambridge OA URLs 0-byte / “too large” failures (Park 2017)
archive’s EPFL infoscience resolver fails for Nat Med + Nat Metab papers (urolithin-a verifier flagged for FEATURE_REQUESTS)

Round 4 propagation done in this pass:

sarcopenia.md Conboy 2005 strain
model-organisms/README.md + _extrapolation-guide.md killifish lifespan
ROADMAP.md updated for all 12 pages

No-action pages (downstream pages don’t yet exist or don’t carry the corrected claim):

TREM2 protein page (not built)
alzheimers-disease (not built)
chronic-inflammation (stub, no CANTOS dose to fix yet)
cellular-senescence + sasp.md (already correct framing for Coppé 2008 timing; CM-specific SASP not duplicated outside cardiomyocytes.md)

🧬 Aging Wiki

Explorer

R4

log/R4.md — Round 4 entries

[2026-05-04] verify | López-Otín 2023 hallmarks paper (PDF acquired via user manual download)

[2026-05-04] add | ROADMAP.md coverage tracker

[2026-05-04] add | wiki-verifier subagent

[2026-05-04] verify | first verification pass against PDFs

[2026-05-04] worked-examples + hallmark stubs

[2026-05-04] update | a local paper archive integration upgraded

[2026-05-04] seed | ampk (wiki-seeder subagent)

[2026-05-04] seed | p53-pathway (wiki-seeder subagent)

[2026-05-04] init | scaffold

[2026-05-04] verify | processes/sasp.md

[2026-05-04] verify | molecules/proteins/atm.md

[2026-05-04] verify | phenotypes/sarcopenia.md

[2026-05-04] verify | pathways/p53-pathway.md

[2026-05-04] verify | processes/mitophagy.md

[2026-05-04] verify | molecules/compounds/quercetin.md

[2026-05-04] verify | molecules/proteins/bcl-xl.md

[2026-05-04] verify | molecules/proteins/raptor.md

[2026-05-04] verify | molecules/proteins/fkbp12.md

[2026-05-04] verify | molecules/proteins/4ebp1.md

[2026-05-04] verify | molecules/proteins/tsc1-tsc2.md

[2026-05-04] verify | molecules/compounds/nmn.md

[2026-05-04] verify | molecules/compounds/nmn.md

[2026-05-04] verify | molecules/compounds/urolithin-a.md

[2026-05-04] ingest | Li 2023 — dual mTOR/NAD+ gerotherapy (bioRxiv preprint)

[2026-05-04] verify | molecules/proteins/bak.md

[2026-05-04] verify | molecules/proteins/mcl-1.md

[2026-05-04] verify | interventions/pharmacological/senomorphics.md

[2026-05-04] verify | model-organisms/drosophila-melanogaster.md

[2026-05-04] verify | model-organisms/caenorhabditis-elegans.md

[2026-05-04] verify | cell-types/cardiomyocytes.md

[2026-05-04] verify | cell-types/microglia.md

[2026-05-04] round-4 summary

Graph View

Table of Contents

Backlinks