🧬 Aging Wiki

R1

22 min read

log/R1.md — Round 1 entries

Sub-file of log — see parent for index.

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

Verified against Zhu 2015 (10.1111/acel.12344), Justice 2019 (10.1016/j.ebiom.2018.12.052), Hickson 2019 (10.1016/j.ebiom.2019.08.069) — all full PDFs read. CAS confirmed via PubChem CID 3062316 synonyms. ChEMBL resolved.

Corrections: (1) chembl-id null → CHEMBL1421; (2) CAS gap marker removed, confirmed; (3) HUVEC D-efficacy “Partial” → “No” per Fig. 2A Zhu 2015; (4) “Murine cardiac fibroblasts” specificity row removed — not assayed in Zhu 2015; (5) INK-ATTAC removed from D+Q treatment models — was genetic comparator only; (6) Justice Q dose 1000 → 1250 mg/day; (7) Justice schedule “D1,D2,D8,D9,D15,D16” → Days 1–3, 8–10, 15–17 (9 total dosing days); (8) Hickson clinical section enriched with confirmed reduction percentages (p16+ −35%, p21+ −17%, SA-βgal+ −62%), macrophage/CLS/skin/SASP/progenitor data. Downstream: quercetin.md and studies/zhu-2015-senolytic-drugs.md may carry inherited errors from the D+Q cell-type table.

[2026-05-04] verify | Round 1 COMPLETE — 12 of 12 pages verified (summary)

All Round 1 priority pages now verified: true (1 full = ampk; 11 verified-partial). The 11-agent parallel batch ran in ~15 min wall-clock vs ~150 min serial. Per-page corrections logged in the entries that follow.

Cross-cutting patterns the verifier batch surfaced:

  • Two inverted claims caught (most serious findings):

    • caloric-restriction.md “TOR-deficient organisms show no further benefit from CR” — Bjedov 2010 actually shows the opposite in flies (rapamycin extends BEYOND DR). Propagated to fix this round.
    • mdm2.md Mendrysa 2006 “accelerated aging” — paper actually shows the mice are tumor-resistant and do NOT age prematurely (110 vs 106 wk, p=0.61). Direct verifier correction.

  • Citation accountability: Multiple cases where the cited paper didn’t actually contain the claimed result. Examples: Freund 2010 doesn’t contain the mTOR-SASP claim it was footnoted for (sasp.md); Levine & Oren 2009 says “hundreds” of p53 targets, not “~500” (p53-pathway.md + p53.md); Lahav 2004 reports 350±160 / 440±100 min pulse intervals, not “~5.5 h” (p53-pathway.md); Eisenberg 2009 didn’t show spermidine lifespan extension in mice (autophagy.md).

  • DOI errors: 5 wrong DOIs corrected — Mendrysa 2006, Lev Bar-Or 2000, el-Deiry 1993 (also 1994→1993 year), Nakano 2001 PUMA, Suh 2008 IGF1R. The seeder relied on training-knowledge DOIs that were wrong; the verifier caught them via Crossref/Europe PMC re-lookup.

  • Mechanism nuances: SASP is p53-INDEPENDENT (ATM/NBS1/CHK2-driven; p53/p21 is the separate growth-arrest arm); AMPK is “direct antagonist” of mTORC1 only via Raptor arm (TSC2 arm is indirect via Rheb-GDP); Mannick 2014 reduced PD-1 (not PD-L1); ULK1 Ser777 is mouse-specific (not conserved in human); cGAS-STING acts in parallel with (not independently of) DDR.

  • Quantitative corrections: many. Examples: mtor.md Drosophila lifespan 10–24% → 7–16%; ampk.md γ-subunit allosteric 2–5× → ~10×; CR Redman 2018 was n=53 sub-study (not n=218 full trial); CR achieved 15% (not 12%); sarcopenia onset ~50 (not ~40); sarcopenia prevalence 1–29% (not “~10–16% canonical”); Shingrix 89.8% (not “>90%”); Holzenberger strain 129/Sv (not C57BL/6).

Propagation pass applied this round (3 cross-page edits):

  • mtor.md — “AMPK direct antagonist” → two-arm framing (Raptor direct, TSC2 indirect via Rheb-GDP)
  • caloric-restriction.md — TOR-epistasis claim corrected to organism-specific (yeast/worms convergence; flies show partial independence per Bjedov 2010)
  • p53.md — “transcriptionally activate ~500 target genes” → “hundreds” + #gap/unsourced + new Vousden-Lane 2007 footnote

Several feared propagations turned out to be unneeded (verifying they didn’t apply saved unnecessary edits):

  • autophagy.md doesn’t claim ULK1 Ser777 specifically (caveat applies on ampk.md only)
  • immunosenescence.md doesn’t claim “restored thymopoiesis” (already says “thymic output itself not clearly restored”)
  • sasp.md table already correctly attributes DDR-driven SASP to ATM/ATR (no p53/p21 conflation present)

Subagent ergonomics observation: The verifier proved exceptionally productive — it flipped 11 pages from verified: false to verified: true with substantive PDF-grounded corrections in ~15 min wall-clock. The 10× parallel speedup makes per-page verification economically viable for routine wiki maintenance. Sonnet model was sufficient for this work.

Outstanding for future lint passes:

  • 4 #gap/unsourced cross-organism magnitude rows in caloric-restriction.md (yeast/worm/fly/rat/mouse lifespan extension figures)
  • Mannick 2014 + several closed-access papers — re-verify if PMC manuscripts become available
  • Yeast/worm CR-TOR convergence claim citation (caloric-restriction.md, mtor.md)
  • TFEB CLEAR network gene-count primary citation (autophagy.md)
  • Mendrysa-related “Mdm2 inhibition causes accelerated aging” claims in any compound/intervention pages — propagation check when those pages are created

(Per-page verifier entries with full details below.)

[2026-05-04] verify | pathways/insulin-igf1.md

Pages verified: 1

  • pathways/insulin-igf1.md — 4 PDFs read end-to-end (Holzenberger 2003, Kenyon 2010, Willcox 2008, Suh 2008); 2 additional papers confirmed closed-access (Kenyon 1993, Taguchi 2007). 8 corrections applied:
    1. Holzenberger 2003 mouse background: “C57BL/6” → “129/Sv” (correct strain per Methods)
    2. Holzenberger 2003 overall lifespan figure: added overall 26% (P<0.02, Cox) previously missing; male figure corrected to 15.9% (NS) from unspecified “ns”
    3. Willcox 2008 n: “n=5345 men” → “n=615 (213 cases, 402 controls)”; 5345 is the source HHP cohort, not the study sample
    4. Willcox 2008 study design: “observational cohort” → “nested case-control”; added OR=2.75 (95% CI 1.51–5.02, P=0.0007)
    5. Suh 2008 unsourced resolved: DOI 10.1073/pnas.0705467105 confirmed; full citation and footnote added; claim now accurately describes n=79 female centenarians screened, two IGF1R variants (Ala-37-Thr, Arg-407-His), P=0.04, functional impairment confirmed in lymphocytes
    6. Magnitude attenuation table: IGF1R+/- row updated with overall 26% figure and corrected male percentage
    7. Taguchi 2007 (~14% lifespan extension): paper is closed-access (not_oa); tagged no-fulltext-access; figure retained but flagged as unverified against source
    8. Kenyon 1993 (daf-2 ~2× lifespan): paper is closed-access (not_oa); claim is well-established and consistent with Kenyon 2010 review confirmation; tagged no-fulltext-access in context

Pages unverifiable (closed-access): Kenyon 1993 (10.1038/366461a0), Taguchi 2007 (10.1126/science.1142179) — both not_oa.

Downstream pages to check: any entity pages citing holzenberger-2003-igf1r-lifespan or willcox-2008-foxo3a for inherited n or strain errors; igf1r if it exists; deregulated-nutrient-sensing if it cites these figures.

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

Pages verified: 1

  • pathways/ampk.md — 4 PDFs read end-to-end; 6 corrections applied:
    1. γ-subunit site 1 allosteric fold-activation: “~2–5 fold” → “up to ~tenfold” (Hardie 2012, p.252)
    2. ~100-fold effect attribution: clarified as Thr172 phosphorylation overall, not site 3 protection alone (Hardie 2012, Fig. 1 caption)
    3. ULK1 Ser777 conservation: added caveat that mouse Ser777 is not conserved in human ULK1 (Kim 2011, Discussion)
    4. Martin-Montalvo lifespan metric: “median” → “mean” lifespan extension (paper reports mean, not median)
    5. Martin-Montalvo quantitative expansion: added χ²=5.46, p=0.02, B6C3F1 replication 4.15% (ns), 1% dose −14.4%, pAMPK +27%
    6. “direct antagonist of mTORC1” framing: qualified in intro and cross-pathway section to distinguish direct (Raptor) vs indirect (TSC2/Rheb) arms

Pages unverifiable (closed-access): none — all 4 DOIs downloaded and read.

Downstream pages to check: metformin compound page (may have inherited “median” error and/or male-only scope omission); autophagy process page (ULK1 Ser777 human conservation caveat); mtor pathway page (“direct antagonist” framing).

[2026-05-04] verify | wiki-verifier subagent confirmed working; Round 1 verification batch dispatched

After session restart, project-level subagents (wiki-verifier, wiki-seeder) register correctly. The Agent tool now accepts subagent_type: wiki-verifier directly.

First verifier run (sasp.md): 8 substantive corrections — cell-line accuracy (Coppé 2008 used 5 named human fibroblast strains, not generic “human and mouse”), mechanism refinements (cGAS-STING acts in parallel with, not independently of, DDR), citation accountability (Freund 2010 doesn’t actually contain the mTOR-SASP claim — replaced with #gap/unsourced and Laberge/Herranz 2015 candidates), confirmed Kang 2015 GATA4 DOI as 10.1126/science.aaa5612 (Science, not Cell as the seeder assumed). The verifier reads PDFs end-to-end and produces meaningfully better scholarship than the seeder’s training-knowledge draft.

Round 1 verification batch dispatched (11 agents in parallel background):

  • pathways/mtor.md
  • pathways/p53-pathway.md
  • pathways/insulin-igf1.md
  • pathways/ampk.md
  • pathways/dna-damage-response.md
  • processes/autophagy.md
  • molecules/proteins/mdm2.md
  • molecules/proteins/atm.md
  • phenotypes/sarcopenia.md
  • phenotypes/immunosenescence.md
  • interventions/lifestyle/caloric-restriction.md

Each prompt provides the seeder-flagged priority items from earlier in this session as focus hints, so verifiers don’t need to rediscover them. Sasp’s lessons (cGAS-STING parallel-not-independent framing; verifier-confirmed Kang 2015 DOI) cross-pointed where relevant (e.g., DDR verifier asked to cross-check the cGAS-STING claim).

Expected wall-clock: ~15 min for the parallel batch (vs ~150 min serial). Each verifier returns a ≤300-word corrections summary. After completions, main thread will: (1) update ROADMAP entries to (verified) or (verified-partial), (2) propagate any cross-page corrections (e.g., a finding on mtor.md that affects autophagy.md), (3) refresh implicit-stub queue.

[2026-05-04] verify | molecules/proteins/mdm2.md

Pages verified: 1 (partial scope)

  • molecules/proteins/mdm2.md — 4 PDFs read end-to-end: Haupt 1997, Kubbutat 1997, Lev Bar-Or 2000, Mendrysa 2006. Vassilev 2004 is closed-access (not_oa); Oliner 1992 download failed (green OA, HTTP 403). 7 corrections applied:
    1. Mendrysa 2006 DOI: 10.1128/MCB.26.17.6372-6382.2006 (does not exist) → 10.1101/gad.1378506 (Genes & Development)
    2. Mendrysa 2006 journal: MCB → Genes & Development
    3. Mendrysa 2006 finding INVERTED: wiki claimed “accelerated aging phenotypes including hematopoietic failure, organ atrophy, and reduced lifespan”; paper actually shows mdm2puro/7-12 mice do NOT age prematurely (lifespan 110 vs 106 wk, log-rank p=0.61; n=28 hypomorphs, n=12 WT). Lymphopenia and intestinal apoptosis are present but are distinct from organismal aging acceleration.
    4. Lev Bar-Or 2000 DOI: 10.1073/pnas.97.21.11250 (not found) → 10.1073/pnas.210171597 (confirmed via Europe PMC; PDF downloaded)
    5. Oscillation period: “~5-hour ultradian oscillations” → “~3 h periodicity” per paper (NIH 3T3 cells, 5 Gy IR)
    6. ATM phosphorylation sites: Ser-240/242/246/260/262 incorrectly attributed to ATM; UniProt Q00987 ATM sites are Ser-386, Ser-395, Ser-407, Thr-419, Ser-425, Ser-429. Corrected in body text; clarification note added for the five non-ATM serines.
    7. Added Ser-386 to frontmatter key-ptms (was omitted).

Pages unverifiable (closed-access or download-failed):

  • Vassilev 2004 (10.1126/science.1092472) — not_oa; DLT and nutlin pharmacology claims tagged in Limitations
  • Oliner 1992 (10.1038/358080a0) — download failed (HTTP 403); cancer amplification frequency tagged in Limitations

Downstream pages to check: p53 (may reference Mendrysa 2006 aging claim); any page citing mdm2 for the hypomorphic aging claim.

[2026-05-04] verify | pathways/dna-damage-response.md

Pages verified: 1 (partial)

  • pathways/dna-damage-response.md — ncb1909 (Rodier 2009), ncb2170 (Di Micco 2011), Jackson & Bartek 2009, Ciccia & Elledge 2010, and Rogakou 1998 JBC all read end-to-end. 5 corrections:
    1. Rodier model: “CHK2 → p53 → p21” conflated growth-arrest arm with SASP arm. Rodier 2009 explicitly shows p53/pRb NOT required for SASP; ATM/NBS1/CHK2 ARE required. Expanded into two separable arms.
    2. SASP trigger: vague “single persistent DSB triggers SASP” → replaced with actual experimental comparison (0.5 Gy transient DDR/no SASP vs 10 Gy PDDF/robust SASP) + ATM/NBS1/CHK2 knockdown evidence.
    3. Paracrine ROS claim: “SASP amplifies DDR via paracrine ROS” not established in Rodier 2009. Replaced with accurate summary + needs-replication.
    4. ncb2170 footnote: “review/primary” → “primary research”; Di Micco R et al. authorship, OIS-specific SAHF content, orphaned-citation note added.
    5. hoeijmakers2009 footnote: added no-fulltext-access (NEJM PDF download failed twice).

Pages unverifiable: Hoeijmakers 2009 NEJM (download failed); Shiloh & Ziv 2013, Zou & Elledge 2003, San Filippo 2008, Lieber 2010 (all not_oa). No cGAS-STING claims on this page.

Downstream pages to check: cellular-senescence.md and sasp.md (Rodier SASP mechanism conflating p53/SASP arm); molecules/proteins/atm.md (ATM→SASP arm).

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

Pages verified: 1 (partial)

  • pathways/mtor.md — Bjedov 2010 and Harrison 2009 PDFs read end-to-end; 5 corrections made:
    1. Drosophila lifespan range “~10–24%” → “~7–16% median lifespan (w^Dah^ females, 50–400 µM, standard food)” — source uses median values in Figure 7A; the upper 24% figure does not appear in the paper at standard food concentrations.
    2. Autophagy mechanism text overstated as “Genetic ablation of ATG genes abolishes most lifespan benefits” → corrected to specific gene (Atg5 knockdown by RNAi) that “completely abolished” extension; 4E-BP identified as a parallel required effector per the paper.
    3. CR-epistasis claim inverted and unsourced — original said “CR longevity effects largely abolished in TOR-deficient organisms”; Bjedov 2010 actually shows the opposite in flies (rapamycin extends BEYOND DR maximum). Replaced with organism-specific nuance and unsourced + needs-replication tags.
    4. Bjedov footnote label “review of multiple cohorts” → corrected to “original research, multiple genetic backgrounds and food concentrations.”
    5. Harrison footnote enriched with actual n (1,960 total, 674 controls), both 90th-percentile (♀ +14%, ♂ +9%) and mean (♀ +13%, ♂ +9%) effect sizes, and caveat about pre-600-day diet differences at UT/UM sites.

Pages unverifiable (closed-access):

  • Mannick 2014 (doi:10.1126/scitranslmed.3009892) — not_oa per a local paper archive; tagged no-fulltext-access on the Mannick claim paragraph. n=218 and “influenza vaccine seroconversion” endpoint NOT verified.

Yeast/worm lifespan rows (~20–40%, ~30–50%) have no primary-source footnotes and were not verified — flagged for future lint pass.

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

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

  • processes/autophagy.md — Hansen 2018 verified via PMC full text (PMC6424591); Eisenberg 2009 verified via PubMed abstract (PMID 19801973); both PDFs failed to download from archive (Hansen 2018: green OA Cambridge URL returned HTML; Eisenberg 2009: bronze OA links returned HTML or HTTP 410). 4 corrections applied:
    1. Spermidine mechanism: “EP300 acetyltransferase inhibitor → relieves cytoplasmic protein acetylation block” → “Histone acetyltransferase (HAT) inhibitor → histone H3 deacetylation → upregulation of autophagy-related transcripts” — Eisenberg 2009 abstract specifies HAT inhibition broadly; EP300 specifically is not named in that paper (EP300 mechanism is from later work, c. 2013, not Eisenberg 2009)
    2. Spermidine lifespan evidence: “Yeast, worms, flies, mice” → “Yeast, worms, flies, human immune cells” — Eisenberg 2009 tested lifespan only in yeast/worms/flies/human cells; mice showed oxidative stress inhibition only, not lifespan extension
    3. Eisenberg 2009 footnote corrected to match above; mechanism note added
    4. TFEB “~500 genes” figure removed — not stated in Hansen 2018 or Settembre 2011; replaced with “many” + unsourced tag pending primary TFEB ChIP-seq citation
    5. AMPK/TSC2/Raptor row in regulation table: TSC2/Raptor detail is not cited to either source paper; tagged unsourced pending AMPK-mTOR primary citation (see pathways/ampk.md verify log for Raptor/TSC2 sourcing)

Pages unverifiable (PDF-level): both DOIs failed to download from archive; verification done against PMC HTML and PubMed abstract. Mechanism table (initiation-stage protein assignments: ATG13, FIP200, ATG101; nucleation: VPS34/VPS15/ATG14L; elongation; cargo receptors; SNARE fusion machinery) verified as standard field knowledge consistent with Hansen 2018 Fig 1 description — not contradicted.

Downstream pages to check: spermidine if it exists — may have inherited the EP300-specific mechanism claim or the mice-lifespan overclaim. spermidine footnote citing Eisenberg 2009 should be updated to match corrected organism list.

[2026-05-04] policy | archive search deprioritized; web-based citation discovery

User policy update: do not use archive search for term-based queries — even if BUG-1 (SQL type-cast errors) is fixed. The a local paper archive is now treated as a DOI-keyed lookup layer only (, ). Term-based citation discovery happens via the web.

Updates applied:

  • feedback_archive_search.md memory saved (cross-session policy)
  • .claude/agents/wiki-seeder.md § Workflow step 3 — replaced archive search recommendation with WebFetch against PubMed eutils / Semantic Scholar / Crossref / Europe PMC; added a 3b “confirm DOIs against archive” step for +
  • sops/retrieving-papers.md § “When searching by topic” — same migration
  • CLAUDE.md § “Working with a local paper archive” — same migration
  • ~/dev/a local paper archive/FEATURE_REQUESTS.md — BUG-1 marked deprioritized with note that the workflow has moved off archive search; original report retained for reference

Practical effect: future seeder invocations should use Semantic Scholar’s API for “find me 5 high-impact papers about X” (returns DOIs + citation counts in one call), then to check whether each is in the local corpus, then if the verifier needs the PDF.

[2026-05-04] seed | Round 1 COMPLETE — 12 of 12 entities drafted

All Round 1 implicit-stub-priority entities drafted across 5 entity types.

Pages added in this session (12 total):

PageTypeLinesModeCitations
pathways/mtor.mdpathway~250main thread3 footnotes
processes/autophagy.mdprocess~200main thread2 footnotes
phenotypes/sarcopenia.mdphenotype~250main thread1 footnote
pathways/p53-pathway.mdpathway—subagent10 (8 verified)
pathways/ampk.mdpathway189subagent4
molecules/proteins/mdm2.mdprotein—subagent4 confirmed + 2 unconfirmed
interventions/lifestyle/caloric-restriction.mdintervention (NEW TYPE)~280main thread4 footnotes
pathways/insulin-igf1.mdpathway203subagent5
pathways/dna-damage-response.mdpathway323subagent10
processes/sasp.mdprocess185subagent6
molecules/proteins/atm.mdprotein160subagent5
phenotypes/immunosenescence.mdphenotype213subagent4

Round 1 seeder batch 3 returns (insulin-igf1, dna-damage-response, sasp, atm, immunosenescence):

  • All 5 pages followed prototype structure cleanly. Total ~1080 lines of new content.
  • 30 primary-source DOIs cited across the batch; majority verified via title-match. 2 mismatches surfaced (Suh 2008 IGF1R prompt accidentally used Willcox 2008 DOI; Rodier ATM-SASP citation in DDR/ATM prompts referenced wrong DOI — ncb2170 vs correct ncb1909). All flagged with #gap/unsourced or #gap/needs-replication.
  • Schema gap surfaced (sasp): process frontmatter has no subtypes: field for inducer-specific variants like MiDAS — currently shoehorned into selective-variants:. Worth considering for CLAUDE.md if this pattern recurs.
  • Schema confirmation: icd-10: null / icd-11: null accepted convention for phenotype pages without diagnostic codes (immunosenescence).

Verifier priority items the seeders surfaced (added to next verification round):

  • p53-pathway: ~20-30 min p53 half-life claim (Haupt 1997); ~500 p53 target genes (Levine & Oren 2009); ~5.5 h pulse period; el-Deiry 1994 + Nakano 2001 DOIs (mismatched titles in archive)
  • ampk: γ-subunit CBS binding-site numbering (Hardie 2012); Raptor Ser722/Ser792 (Gwinn 2008); ULK1 Ser317/Ser777 (Kim 2011); Martin-Montalvo 2013 lifespan extension %
  • mdm2: ATM phosphorylation site list vs UniProt Q00987; chase Mendrysa 2006 + Lev Bar-Or 2000 DOIs (not in archive)
  • insulin-igf1: Holzenberger 2003 IGF1R+/- lifespan numbers (PDF available); Kenyon 2010 cross-organism extension figures; find correct Suh 2008 IGF1R centenarian DOI
  • dna-damage-response: Rodier 2009 (ncb1909) persistent foci → SASP claims (PDF downloaded); Hoeijmakers 2009 NEJM progeroid table
  • sasp: Coppé 2008 composition + cell-type specificity claims; Wiley 2016 MiDAS quantitative claims; Glück 2017 cGAS-STING mechanism; chase the GATA4/autophagy-SASP citation (Kang 2015)
  • atm: Bakkenist 2003 Ser1981 autophosphorylation mechanism (PDF downloaded); Rodier 2009 ATM-inhibition abrogates SASP
  • immunosenescence: thymic output ~95% decline by 70; CMV-specific CD8+ 10-50% of pool; influenza vaccine seroprotection 40-60% in 65+; Fluzone HD ~24% advantage; IRP cutpoints from primary OCTO/NONA papers (not in archive)

Implicit-stub queue refresh needed — the ~1080 new lines added many new wikilinks. Next lint should rerun the discovery script in sops/lint-pass.md Step 3 to refresh ROADMAP. Notable new high-demand items expected: ATR, BRCA1, CHK2, FOXO3, NF-κB, mitophagy, rapamycin, metformin, NMN, NR.

Round 1 ROADMAP status:

  • 12 of 12 priority entities drafted ([x] (drafted))
  • 0 verified (next round of wiki-verifier invocations needed — many PDFs already downloaded by seeders)
  • Round 2 ready to start: senolytic family completion (quercetin, dasatinib, navitoclax) + Tier 2 implicit stubs (mitophagy, hematopoietic-stem-cells, bcl-xl, cbp-p300, p21, apoptosis-pathway, pi3k-akt-pathway, sirtuin)

Subagent ergonomics observation: Each parallel seeder cost ~5 min wall-clock (running concurrently). 5-page batch returned in ~5 min total — vs ~25 min if serial. Seeder fallback to general-purpose + wiki-seeder.md as system prompt works but verbose; project-level wiki-seeder agent definition will register on next session restart.

[2026-05-04] verify | interventions/lifestyle/caloric-restriction.md

Pages verified: 1 (partial — McCay 1935 unverifiable, closed-access)

PDFs read: Mattison 2017 (ncomms14063, 12 pp), Redman 2018 (cmet.2018.02.019, 11 pp), Spadaro 2022 (science.abg7292, 5 pp). McCay 1935 (jn/10.1.63): not_oa, cannot verify.

Corrections applied (8):

  1. Redman 2018 n: “~220 humans” → “n=53 (34 CR, 19 AL control)” — ~220 is full CALERIE II enrolment; Redman 2018 is Pennington ancillary metabolic-chamber sub-study.
  2. Achieved CR %: “~12%” → “~15% (14.8% ± 1.5%)” per Redman 2018 Fig. 2A; Spadaro 2022 reports ~14%; full trial average ~12% (Ravussin 2015).
  3. Redman 2018 primary findings: removed misattributed “LDL, blood pressure, insulin sensitivity” bullet (Martin 2016, not Redman 2018); corrected to SleepEE metabolic adaptation (~7% beyond weight loss, p<0.01) and reduced F2-isoprostane (p<0.01). Safety outcomes (bone/mood/cognition) also moved to Martin 2016.
  4. PLA2G7 attribution: moved from Redman 2018 to Spadaro 2022 (correct source).
  5. Spadaro 2022 thymopoiesis language: “restored thymopoiesis” → “improved thymopoiesis” — paper does not claim full restoration.
  6. Mattison 2017 magnitude: removed fabricated “~5–10% lifespan extension” figure; replaced with actual HR 1.865 (95% CI 1.119–3.108, p=0.017) for UW adult-onset only.
  7. Mattison 2017 cross-table: clarified UW-significant / NIA-not-significant split vs misleading “joint reanalysis showing significant extension across sites.”
  8. Cross-organism magnitude ranges (yeast/worm/fly/rat/mouse) tagged unsourced — no primary-source footnotes on page.

Footnotes rewritten with verified n, design, endpoints. McCay 1935 tagged no-fulltext-access.

Downstream pages to check:

  • phenotypes/immunosenescence.md — likely inherited “restored thymopoiesis” language from Spadaro 2022; should read “improved.”
  • Any page citing CALERIE II n=~220 as Redman 2018’s n (vs full trial enrolment).

[2026-05-04] seed | Round 1 batch 2 — first parallel seeder run + caloric-restriction prototype

Subagent batch (3 in parallel; general-purpose with wiki-seeder.md as effective system prompt — project-level subagents still aren’t registered in this session):

  • pathways/p53-pathway.md (drafted, verified: false) — DDR upstream → p53 stabilization → 3 downstream programs (cell-cycle arrest via p21, senescence, apoptosis via PUMA/BAX/NOXA). Cross-links to existing [[p53]] protein page; avoids duplicating its content. KEGG hsa04115 + Reactome R-HSA-3700989 / R-HSA-5633007 confirmed via WebFetch. Seeder cited 10 DOIs (8 verified via title-match; 2 mismatched — el-Deiry 1994 + Nakano 2001 — flagged for verifier).

  • pathways/ampk.md (drafted, verified: false) — heterotrimeric kinase structure (γ-subunit AMP/ADP/ATP allosteric switch); LKB1 + CaMKK2 upstream; mTORC1/TSC2 + Raptor + ULK1 + ACC + SREBP1 + PGC-1α + FOXO downstream; cross-organism longevity evidence with extrapolation rubric. KEGG hsa04152 + Reactome R-HSA-380972 + WikiPathways WP1471 confirmed. Cited 4 DOIs (Hardie 2012, Kim 2011, Gwinn 2008, Martin-Montalvo 2013), all title-verified.

  • molecules/proteins/mdm2.md (drafted, verified: false) — primary p53 negative regulator (E3 ubiquitin ligase); MDM2-p53 negative feedback loop; ATM-mediated stress disruption; ~7% cancer amplification; pharmacology (nutlin-3, RG7388, AMG-232, DLT thrombocytopenia). UniProt Q00987 + NCBI Gene 4193 pulled via REST API. Cited 4 confirmed DOIs (Haupt 1997, Kubbutat 1997, Oliner 1992, Vassilev 2004) + 2 unconfirmed (Lev Bar-Or 2000, Mendrysa 2006) tagged #gap/needs-replication.

Main-thread (caloric-restriction is a new entity-type prototype):

  • interventions/lifestyle/caloric-restriction.md (drafted, verified: false) — establishes the intervention page schema (type: intervention; new optional fields: mode, mechanisms, target-hallmarks, target-pathways, human-evidence-level, clinical-stage, safety-profile). Sections: terminology disambiguation (CR / DR / IF / TRE / MetR / FMD), mechanisms with mTOR/AMPK/IIS/autophagy/sirtuin cross-links, cross-organism evidence table including the Mattison 2017 primate reanalysis and CALERIE II thymopoiesis finding, side effects + limitations, CR mimetics list, cross-organism extrapolation rubric, key open questions.

Schema additions (now in CLAUDE.md):

  • New type: intervention page-type schema. Convention documented: interventions/lifestyle/ for direct lifestyle pages; interventions/pharmacological/ for drug-class category pages only — specific drugs live as compounds in molecules/compounds/.

Infrastructure issues surfaced (filed against a local paper archive):

  • BUG-1: archive search PostgreSQL type-cast errors — blocks wiki-seeder citation discovery. Independent confirmation by 2 sonnet subagents (specific SQL errors quoted). Filed as BUG-1 in a local paper archive/FEATURE_REQUESTS.md.
  • BUG-2: DOI-to-title mismatches in for some records (el-Deiry 1994, Nakano 2001 specifically flagged). Filed as BUG-2. Wiki agents now cross-check titles against context before trusting metadata.

Verifier priority items the seeders surfaced (for next verification round):

  • p53-pathway: ~20–30 min p53 half-life claim (Haupt 1997); ~500 p53 target genes figure (Levine & Oren 2009); ~5.5 h pulse period claim
  • ampk: γ-subunit CBS domain binding-site numbering + fold-activation estimates (Hardie 2012); Raptor phosphorylation sites Ser722/Ser792 (Gwinn 2008); ULK1 phosphorylation sites Ser317/Ser777 (Kim 2011); Martin-Montalvo 2013 lifespan extension % + dose
  • mdm2: cross-check ATM phosphorylation site list against UniProt Q00987 directly; chase Mendrysa 2006 + Lev Bar-Or 2000 DOIs (not in archive)

Next: 5 more seeders launched in background — insulin-igf1, dna-damage-response, sasp, atm, immunosenescence — completing Round 1’s pathway / process / protein / phenotype demand. Will synthesize when notifications return.

ROADMAP.md updated: p53-pathway / ampk / mdm2 / caloric-restriction flipped from [ ] to [x] (drafted).

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

Pages verified: 1 (partial — Nikolich-Žugich 2018 and Franceschi & Campisi 2014 unverifiable)

PDFs/texts read: Goronzy & Weyand 2019 (PMC7584388 HTML full text, green OA but PDF download failed in archive); Mannick 2014 (published abstract via PubMed and Europe PMC); DiazGranados 2014 NEJM (PubMed abstract, PMID 25119609); Cunningham 2016 NEJM ZOE-70 (Europe PMC abstract). Nikolich-Žugich 2018 (doi:10.1038/s41590-017-0006-x): not_oa. Franceschi & Campisi 2014 (doi:10.1093/gerona/glu057): not_oa.

Corrections applied (5):

  1. PD-L1 → PD-1: The wiki stated Mannick 2014 “reduced PD-L1 expression on T cells.” The published abstract explicitly says “reduced the percentage of CD4 and CD8 T lymphocytes expressing the programmed death-1 (PD-1) receptor.” PD-L1 (ligand) and PD-1 (receptor on T cells) are distinct; this was a meaningful factual error. Corrected in body text and footnote.
  2. Mannick ~20% confirmed: Abstract states “RAD001 enhanced the response to the influenza vaccine by about 20%.” Wiki description now includes this figure explicitly. n=218 remains unverifiable from abstract (full text is not_oa).
  3. Shingrix efficacy: “>90% in 70+” → “89.8% in 70+” per ZOE-70 trial (Europe PMC; 95% CI 84.2–93.7; n=13,900). The original “>90%” slightly overstated the data.
  4. Fluzone HD efficacy: “~24%” confirmed as 24.2% (95% CI 9.7–36.5; DiazGranados 2014 NEJM, n=31,989 phase IIIb-IV RCT). Exact figure and CI now in table.
  5. Thymic output framing: Goronzy & Weyand 2019 states thymic contribution to T cell generation declines from ~16% to <1% over adult lifetime — a different metric from the wiki’s “~95% decline in thymic output by age 70.” Both claims tagged with no-fulltext-access (Nikolich 2018 source) and explanatory note added. The ~95% figure was not contradicted but could not be sourced.

Gap tags added:

IRP attribution clarified: primary Wikby et al. 1998 OCTO paper (PMID 9720651) named as original source; Nikolich 2018 is a secondary review. Cutpoints require verification against Wikby primary papers.

Pages unverifiable (closed-access):

  • Nikolich-Žugich 2018 (doi:10.1038/s41590-017-0006-x) — not_oa; 3 quantitative claims depend on this source
  • Franceschi & Campisi 2014 (doi:10.1093/gerona/glu057) — not_oa; inflammaging-immunosenescence distinction paragraph depends on this source
  • Mannick 2014 (doi:10.1126/scitranslmed.3009892) — not_oa; n=218 and dose data unverified (abstract only)

Downstream pages to check:

  • mtor — cites Mannick 2014 and may have inherited the PD-L1 error; check for “PD-L1” and correct to “PD-1”
  • Any page citing the ~95% thymic output figure should be tagged if inherited from this page

[2026-05-04] seed | Round 1 main-thread exemplars + wiki-seeder definition

Drafted three first-of-type pages in main thread to establish entity-type conventions:

  • pathways/mtor.md — pathway prototype. Sections: complex biology (mTORC1 vs mTORC2), upstream/downstream tables, role-in-aging cross-organism evidence table, pharmacology by drug class, footnotes citing Bjedov 2010 / Harrison 2009 / Mannick 2014.
  • processes/autophagy.md — process prototype. Sections: stage-by-stage mechanism table (initiation → fusion), regulation table, selective autophagy variants, intervention table, methods/readout caveats. Cited Hansen 2018 / Eisenberg 2009.
  • phenotypes/sarcopenia.md — phenotype prototype. Sections: competing diagnostic criteria (EWGSOP2 / AWGS / FNIH), pathophysiology grouped by mechanism, risk-factor table, intervention evidence ranking, hallmark-mapping section.

All three ship with verified: false + auto-extraction banner. Verifier subagent will cross-check against primary sources later.

Schema additions made during drafting (now reflected in CLAUDE.md):

  • Pathway: optional wikipathways: ID, optional sens-categories: mapping
  • Process: optional selective-variants: field (for processes with sub-variants like autophagy → mitophagy)
  • Phenotype: optional icd-11:, optional prevalence-65plus:
  • New convention note: pathway/protein naming clash resolution (pathway gets bare name, protein uses suffix like -kinase when seeded later)

Created wiki-seeder subagent definition at .claude/agents/wiki-seeder.md. Tools: Read, Write, Edit, Bash, WebFetch. Model: sonnet. Workflow: load CLAUDE.md + relevant SOP + entity-type prototype, pull from canonical DBs, discover citations via a local paper archive, draft per schema with verified: false banner. Hard rules: no fabricated citations, no invented canonical IDs, escalate schema gaps explicitly, don’t propagate downstream.

ROADMAP.md updated: mtor / autophagy / sarcopenia flipped from [ ] to [x] (drafted).

Pairs with wiki-verifier: seeder drafts → verifier flips verified: true after PDF cross-check → main agent propagates downstream.

Next: spawn 3 seeder subagents in parallel for the next batch of Round 1 (one per established entity type) to validate the parallel-pages workflow before scaling.

Graph View

Backlinks