log/R31.md — Round 31 entries

Sub-file of log — see parent for index.

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

Pages verified: 1 (partial — 3 of 7 cited sources verified against local PDFs; 2 closed-access unverifiable; 2 review PDFs failed download)

molecules/proteins/atr.md — 5 corrections to Murga 2009 and Ruzankina 2007 claims; UniProt/GenAge/Open Targets canonical fields confirmed

Sources checked against local PDFs:

doi:10.1038/ng.420 (Murga 2009, ATR-Seckel mouse) — local PDF read in full; 3 corrections
doi:10.1016/j.stem.2007.03.002 (Ruzankina 2007, conditional Atr deletion) — downloaded (bronze OA, camoufox) and read in full; 2 corrections
doi:10.1126/science.1083430 (Zou & Elledge 2003) — cross-checked from verified rpa page per R32b SOP; no re-read needed

Canonical-DB IDs verified:

UniProt Q13535 — confirmed: 2,644 aa, Swiss-Prot (reviewed) — CORRECT
Open Targets ENSG00000175054 — confirmed: Advanced Clinical SM (no Approved Drug for ATR); druggability tier 2 confirmed via aging-context convention — CORRECT
GenAge entry 231 — confirmed as ATR (human gene) via HAGR web — CORRECT

Sources unverifiable:

doi:10.1073/pnas.93.7.2850 (Cimprich 1996) — not_oa; no-fulltext-access; cloning-only, low impact on ATR aging claims
doi:10.1042/bj20102162 (Nam & Cortez 2011) — PMC XML accessed (PMC3678388); archive PDF download failed; quantitative activation fold-change not independently verifiable from this review
doi:10.1038/nrm.2017.67 (Saldivar 2017) — DOI lookup failed; claims attributed to this review unverified
doi:10.1126/science.1140321 (Matsuoka 2007) — closed-access; substrate count (“>700 proteins”) not verifiable against full text

Corrections:

“intestinal crypt hypoplasia, lymphopenia” (Murga 2009-attributed) → replaced with paper-accurate “pancytopenia (decreased RBC, WBC, platelets)” + hair/kyphosis/osteoporosis/fat accumulation details; intestinal hypoplasia is Ruzankina 2007 not Murga Seckel mice
“ATR-Seckel mice die at an average of ~6 months” → “less than half a year” (paper’s own language); KM curve n=20 Atr^+/+ and n=27 ATR^S/S; median ~20 weeks; “~6 months” overstated paper
Ruzankina 2007 body: “intestinal failure” → “acute intestinal villus epithelial loss (~80% at 1 week) recovering by 1 month; permanent failures in thymus, skin, testes, bone marrow” — recovery detail from Ruzankina Fig 4A; added quantitative aging data (20% weight loss, 18% cortical bone decrease, 46–76% trabecular bone volume reduction)
Ruzankina 2007 footnote: updated from “PDF download pending” to “downloaded 2026-05-07, verified”; strain clarified (129SvEv/C57BL/6 mixed)
TOPBP1 “>10-fold” activation: tagged needs-primary-source — Nam & Cortez 2011 is a review; this value originates from primary biochemical studies summarized therein; full PDF not accessible

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

Pages verified: 1 (partial — Lee & Paull 2005 not_oa, unverifiable)

molecules/proteins/mrn-complex.md — 10 corrections; 6 of 8 cited sources verified against local PDFs; 1 closed-access (Lee & Paull 2005); 1 closed-access Petrini 1995 (cloning-only, low impact)

Sources checked (verified against local PDFs):

doi:10.1016/s0092-8674(00)81175-7 (Carney 1998) — local PDF; read in full
doi:10.1038/nature00922 (Hopfner 2002) — local PDF; read in full
doi:10.1016/s0092-8674(00)81547-0 (Stewart 1999) — downloaded and read in full
doi:10.1016/s1097-2765(00)80097-0 (Paull & Gellert 1998) — downloaded and read in full
doi:10.1101/gad.13.10.1276 (Paull & Gellert 1999) — downloaded and read in full
doi:10.3390/biom5042877 (Lavin 2015) — downloaded and read in full

Sources unverifiable:

doi:10.1126/science.1108297 (Lee & Paull 2005) — not_oa, no PMC deposit; no-fulltext-access; cross-checked via Lavin 2015 + atm (verified); consistent
doi:10.1006/geno.1995.1217 (Petrini 1995) — not_oa; no-fulltext-access; cloning-level claims only, low-impact on factual content

Canonical-DB IDs verified:

UniProt P49959 (MRE11), Q92878 (RAD50), O60934 (NBS1) — confirmed via REST API; lengths match
NCBI Gene 4361 (MRE11), 10111 (RAD50), 4683 (NBN) — confirmed via NCBI efetch; gap tags removed
GenAge: null confirmed for all three human subunits

Corrections:

“within seconds of break formation” → “within 30 min” (Carney 1998 cites Nelms et al. 1998 for 30-min timing; “seconds” was unsourced)
MRN intro and Step 1 description both corrected to “within 30 min” with citation
NBS phenotype list: removed unsourced items “growth retardation (postnatal onset)” and “hypogammaglobulinemia”; replaced with source-accurate feature list from Carney 1998 (microcephaly, immunodeficiency, hematopoietic malignancy, radiosensitivity, chromosome instability)
Carney 1998 NBS finding corrected: p95 is “absent” in all 5 NBS cell lines (not “absent or truncated”); hMRE11/hRAD50 levels remain normal — added this mechanistically important detail
ATLD section: expanded with specific mutation details from Stewart 1999 (ATLD1/2: nt1897 C→T stop at R633; ATLD3/4: nt350 A→G N117S missense); added distinguishing features (no telangiectasia, no immunodeficiency, no cancer); removed unsupported “longer survival” phrasing
MRE11 nuclease description: removed unsourced “~20 nt internal incision bypassing protein adducts” claim; corrected to what Paull & Gellert 1998 actually shows (3’→5’ exonuclease + hairpin endonuclease; 3–4x activity increase in Mre11/Rad50 complex)
NBS1 stimulation description: replaced vague “potentiates endonuclease cleavage” with specific Paull & Gellert 1999 findings (≥60-fold hairpin cleavage increase; ATP-induced switch in endonuclease specificity; Rad50 responsible for ATP binding)
RAD50 coiled-coil arm length: “~500 Å each” → “150–600 Å per arm” (exact range from Hopfner 2002 text)
Zinc-hook radiation sensitivity: added quantitative detail from Hopfner 2002 (C1G: 5-fold increase at 300 Gy; C2G: 95-fold increase at 300 Gy)
Hopfner 2002 footnote: corrected PDF path from (stale local path) to (correct verified path)

Other updates:

All pending-status footnotes updated to locally-available with correct paths
gap/needs-canonical-id tags for NCBI Gene IDs removed (verified)
gap/no-fulltext-access tag added for Lee & Paull 2005 in Limitations section
verified flag flipped to true with partial-verification scope note

Downstream pages that may need updates:

None identified — mrn-complex.md is a newly seeded page; no other wiki pages yet cite it via footnote

[2026-05-07] verify — interventions/pharmacological/dna-pkcs-inhibitors.md

Pages verified: 1

interventions/pharmacological/dna-pkcs-inhibitors.md — 8 corrections; 3 of 8 cited sources verified against primary PDFs or full-text confirmation

Sources checked (verified):

doi:10.1016/j.cmet.2017.04.008 (Park 2017) — local PDF; read in full
doi:10.1038/s41416-020-01151-6 (van Bussel 2020) — downloaded and read in full (PMC OA)
PMID:14522929 (Veuger 2003) — confirmed via PubMed efetch (no DOI indexed)

Sources unverifiable:

doi:10.1016/j.bmcl.2004.09.060 (Leahy 2004) — download failed (no OA URL); no-fulltext-access
doi:10.1002/glia.70028 (Rim 2025) — download failed (no OA URL); no-fulltext-access
doi:10.1016/j.dnarep.2010.07.012 (Lowndes 2010) — not_oa; no-fulltext-access
doi:10.1038/301527a0 (Bosma 1983) — not_oa; no-fulltext-access
doi:10.1016/j.apsb.2021.07.029 (Wang 2021) — PDF downloaded but not read this pass (qualitative-only claim, low priority)

Corrections (Park 2017):

NU7441 dose “5 mg/kg i.p.” → REMOVED (fabricated; paper uses oral delivery in HFD pellets; no mg/kg i.p. dose stated); replaced with accurate delivery description
“n=not-specified” → updated to per-experiment n’s from figure legends (n=10/group gamma-H2AX; n=5/group rhesus p-DNA-PK; n=6/group rhesus p-HSP90α; n=5/group mouse p-HSP90α; n=3–6/genotype SCID)
“~3-fold elevated in aged muscle” → clarified as from rhesus p-DNA-PK data (Fig 1B, 15-yr vs 1-yr, Mann-Whitney p<0.01); note added that direct p-DNA-PK measurement was not feasible in mice (antibody cross-reactivity limitation stated in paper)
Body prose and footnote now accurately specify NU7441 route (oral in food pellets) and note 1 µM as the in-vitro dose
Park 2017 footnote: n’s added, route corrected, per-experiment details added

Corrections (van Bussel 2020): 6. “MTD 400 mg once daily” → CORRECTED: MTD was not reached; RP2D declared as 400 mg BID (not QD) 7. “DLTs: thrombocytopenia (Grade 3), fatigue” → CORRECTED: one DLT occurred at 300 mg BID (GI AEs + long recovery); thrombocytopenia was NOT a DLT in this paper; fatigue was a TEAE not a DLT 8. “confirmed partial responses in 3/21 evaluable; median duration 7.6 weeks” → CORRECTED: zero partial or complete responses; 12/31 (39%) stable disease; median treatment duration 6.0 (0.3–69.0) weeks 9. n: clarified as 31 enrolled (Safety Set) / 21 (Efficacy/Dose-Escalation Set); wiki previously stated n=21 without qualification

Veuger 2003 probe identity confirmed: NU7026 was the probe used (not NU7441). Wiki correctly distinguished the two compounds in the class table — no correction needed.

Schema items confirmed:

mechanisms: [dna-pk-inhibition, nhej-blockade, radiosensitization] — consistent with intervention-classes.md R33 entry ✓
target-hallmarks: ["[[genomic-instability]]"] — confirmed ✓
translation-gap: translation-blocked-safety — appropriate; SCID safety barrier is the primary structural blocker ✓
literature-checked-through: 2026-05-07 — confirmed in frontmatter ✓
clinical-trials-active: 4 — 4 active oncology trials confirmed; aging-indication count 0 ✓
dna-pk-inhibition and nhej-blockade added to intervention-classes.md under R33 ✓

Downstream propagation needed:

studies/park-2017-dnapk-ampk-aging.md (if exists/when seeded): study page should reflect corrected n’s, NU7441 delivery route, and rhesus-specific nature of p-DNA-PK 3-fold elevation
studies/vanbussel-2020-peposertib-phase1.md (if exists/when seeded): must reflect MTD not reached, RP2D=400 mg BID, zero partial responses, n=31 enrolled, median treatment 6.0 weeks
molecules/proteins/dna-pkcs.md (verified): cross-check whether the NU7441 “5 mg/kg i.p.” dose appears on that page — if so, correct there as well

[2026-05-07] verify — molecules/proteins/mlh1.md

Pages verified: 1

molecules/proteins/mlh1.md — 7 corrections; 4 of 6 cited sources verified against primary PDFs; 1 closed-access (Baker 1996); 1 cross-checked against verified pathway page (Le 2015)

Sources checked (verified):

doi:10.1038/368258a0 (Bronner 1994) — local PDF
doi:10.1038/ng931 (Lipkin 2002) — local PDF
doi:10.1073/pnas.95.12.6870 (Herman 1998) — downloaded via PMC (PMC22665)
doi:10.1056/NEJMoa064522 (Hitchins 2007) — downloaded OA

Sources unverifiable:

doi:10.1038/ng0796-336 (Baker 1996) — not_oa; MLH1-null mouse claims tagged no-fulltext-access

Sources cross-checked without PDF re-read:

doi:10.1056/NEJMoa1500596 (Le 2015) — cross-checked against verified mismatch-repair page per SOP §3; figures match (n=41, ORR ~40% vs 0%, PFS 78% vs 11%)

Canonical database verification:

UniProt P40692: 756 aa confirmed via REST API
Open Targets ENSG00000076242: tier 4 confirmed (no approved drug SM/AB/PR)
GenAge 243: API endpoint returned HTML (not JSON); not independently confirmed — tag for lint pass

Corrections:

~50% Lynch proportion misattributed to Bronner 1994 → corrected with note that Bronner 1994 does NOT make this claim; proportion is from epidemiologic reviews (Lynch 2009); added needs-replication qualifier
MutLγ “backup MMR endonuclease at base-base mismatches” [Lipkin 2002] → corrected; Lipkin 2002 explicitly shows Mlh3-null mice have <1% MSI in fibroblasts — Mlh3 is NOT required for somatic MMR; meiotic phenotype confirmed (viable but sterile; spermatocyte apoptosis at metaphase; 7% oocyte 2-cell progression)
“~15% of sporadic CRCs are MSI-H” → corrected to “~13–15%”; Herman 1998 says ~13%; Hitchins 2007 intro says ~15% (citing Herman); distinction noted in text
Herman 1998 primary quantitative findings added: 11/13 (84%) MSI+ sporadic CRC methylated; 0/34 MSI-H MSH2 methylation; functional MMR restoration in 3 cell lines
Hitchins 2007 case count added: 24 screened → 2 probands → 1 confirmed mother-to-son transmission; 42% PBL methylation vs 0% sperm methylation quantified; 3 siblings with haplotype but no epimutation noted
Bronner 1994 cloning details added: 756 aa, 3p21.3-23 FISH localization, Ser→Phe at aa 44 mutation in family 2
Lipkin 2002 footnote updated: sterility phenotype, meiotic arrest stages, oocyte 7% 2-cell progression, <1% MSI finding, BAC transgene rescue
⚠️ auto-extraction banner removed

Downstream propagation needed:

pathways/mismatch-repair.md — already verified; MutLγ MMR claim is more nuanced there (consistent with correction)
molecules/proteins/mlh3.md — if seeded, the “backup MMR” role of MLH3 should be corrected per Lipkin 2002 finding

[2026-05-07] verify — interventions/pharmacological/parp-inhibitors.md

Pages verified: 1

interventions/pharmacological/parp-inhibitors.md — 7 corrections; 7 of 10 cited sources verified against primary PDFs

Sources checked (verified):

doi:10.1038/nature03443 (Bryant 2005) — local PDF
doi:10.1038/nature03445 (Farmer 2005) — local PDF
doi:10.1016/j.cell.2013.06.016 (Mouchiroud 2013) — local PDF
doi:10.1007/s11357-025-01679-6 (Kieronska-Rudek 2025) — downloaded OA hybrid
doi:10.1038/s41375-023-02040-6 (Nuttall Musson 2024) — downloaded OA hybrid
doi:10.1126/science.aam7344 (Lord & Ashworth 2017) — PMC author manuscript
doi:10.1056/NEJMoa1706450 (Robson 2017 OlympiAD) — downloaded OA

Sources unverifiable:

doi:10.1126/scitranslmed.aaf9246 (Pommier 2016) — not_oa; no-fulltext-access
doi:10.1002/pros.24712 (Marshall 2024) — not_oa; no-fulltext-access
doi:10.1016/j.cell.2014.03.026 (Fang 2014) — download pending; not verified

Corrections:

Bryant 2005: “>10,000-fold selectivity” removed — not stated in source; Lord 2017 says “~1,000-fold” for the class; footnote replaced with actual experimental details
Mouchiroud 2013: PJ34 → AZD2281 (olaparib, 100 nM) + ABT-888 (veliparib, 100 nM); lifespan extensions +22.9% / +15% confirmed; n=60–100 C. elegans per condition confirmed
Nuttall Musson 2024: n “not-specified” → n=6; cancer type “prostate” → ovarian; footnote + body corrected
Kieronska-Rudek 2025: n=3–6 per assay confirmed; cell model corrected (RAW264.7 only, not BMDM); SA-β-gal and cell-death proportions quantified
Robson 2017: OS HR (0.90, p=0.57) and randomization ratio (2:1) added; statistical method noted
SOLO-2 table entry: flagged as unverified against primary source (Pujade-Lauraine 2017)
Removed ⚠️ auto-extraction banner

Downstream propagation needed:

molecules/proteins/parp1.md — check whether PJ34 appears there for Mouchiroud 2013 citations; correct to AZD2281/ABT-888 if so
studies/mouchiroud-2013-nad-sirtuin-longevity.md — if seeded, update PARP inhibitor attribution

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

Pages verified: 1

molecules/proteins/rpa.md — 9 corrections; 3 of 5 cited sources verified from PDFs/full text; 2 remain closed-access (Wold 1997, Zou & Elledge 2003); canonical database IDs fully confirmed via API

Sources checked:

doi:10.1093/emboj/21.7.1855 (Bochkareva 2002, EMBO J) — downloaded bronze OA via Europe PMC + verified (2.8 Å crystal structure)
doi:10.1016/j.cell.2013.10.043 (Toledo 2013, Cell) — downloaded bronze OA via camoufox + verified
doi:10.1016/j.dnarep.2014.04.008 (Ashley 2014, DNA Repair) — PDF download failed but PMC full text (PMC4135522) verified via NCBI efetch
doi:10.1146/annurev.biochem.66.1.61 (Wold 1997) — closed-access (not_oa); claims remain AI-extracted; tagged no-fulltext-access
doi:10.1126/science.1083430 (Zou & Elledge 2003) — closed-access (not_oa); claims remain AI-extracted; tagged no-fulltext-access
UniProt P27694/P15927/P35244 verified via REST API; NCBI Gene 6117/6118/6119 confirmed; HGNC 10289/10290/10291 confirmed
GenAge HAGRID:67 (human RPA1) confirmed via genenames.org HGNC entry lookup

Corrections:

DBD-F boundary: “residues ~1–120” → “residues ~1–110” (Bochkareva 2002 Fig. 1A: RPA70N = aa 1–110)
DBD-B boundary: “residues ~300–420” → “residues ~301–422” (Bochkareva 2002 Fig. 1A)
DBD-C boundary: not previously stated — added “residues ~436–616” with zinc-ribbon motif detail (Bochkareva 2002)
RPA32 domain residues: added DBD-D (~43–171) and C-terminal module (~200–270) boundaries from Bochkareva 2002 Fig. 1A
Trimerization mechanism: corrected from “three-way OB-fold interface” → three parallel C-terminal α-helices forming a hydrophobic bundle (Bochkareva 2002)
ssDNA binding: corrected from a single-step “~30 nt Kd low nM” description → multistep model (8–10 nt initial via DBD-A/B → 30 nt extended via all four DBDs) per Bochkareva 2002
Toledo 2013 cell lines: “U2OS + HCT116” → “U-2-OS (primary), DLD1-ATR-WT, DLD1-ATR-Seckel, BJ primary fibroblasts” — HCT116 was never used in this paper
Toledo 2013 stressor: “aphidicolin-induced fork stalling” → “hydroxyurea (HU) as primary stressor; aphidicolin and others used for generality validation”
Ashley 2014 phospho-site list: Ser-12 was omitted — added to the cluster (Ser-4, Ser-8, Ser-12, Thr-21, Ser-23, Ser-29, Ser-33); Thr-21 kinase corrected from “ATM” to “one or more PIKKs depending on stress agent”
HGNC IDs for RPA2 (10290) and RPA3 (10291): gap resolved, confirmed via UniProt REST — added to subunit table

Downstream pages to check (main agent task):

molecules/proteins/brca2.md — cites RPA displacement by BRCA2; consistent with wiki claims (no correction needed for this claim)
pathways/homologous-recombination.md — may inherit RPA structural/mechanism claims
pathways/nucleotide-excision-repair.md — may inherit RPA NER-gap-coating claims
molecules/proteins/ku70-ku80.md — cross-reference for NHEJ/HR competition

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

Pages verified: 1

molecules/proteins/xpf.md — 9 corrections; all 7 cited sources verified against PDFs (4 downloaded during this pass)

Sources checked:

doi:10.1016/s0092-8674(00)80155-5 (Sijbers 1996, Cell) — local PDF verified
doi:10.1038/nature05456 (Niedernhofer 2006, Nature) — local PDF verified; KEY aging paper
doi:10.1038/nature19329 (Vermeij 2016, Nature) — local PDF verified
doi:10.1038/ncb1866 (Garinis 2009, Nat Cell Biol) — downloaded OA via PMC + verified
doi:10.1111/acel.13302 (Birkisdottir 2021, Aging Cell) — downloaded OA via PMC + verified
doi:10.1038/sj.emboj.7601894 (Tsodikov 2007, EMBO J) — downloaded OA via Europe PMC + verified
doi:10.1093/nar/gkz769 (Zhang 2019, NAR) — downloaded OA via PMC + verified
UniProt Q92889 verified via REST API; GenAge HAGRID 261 confirmed via web query

Corrections:

Footnote key [^park2019] → [^tsodikov2007] (paper year is 2007; all inline citations updated)
Author attribution: “Park CJ, Choi BS” → “Tsodikov OV, Ivanov D, Orelli B et al.” — entire author list was wrong in auto-extraction
Active-site residue source: Asp-687/Glu-689 attributed to Sijbers 1996 (incorrect) → attributed to Enzlin & Schärer 2002; UniProt Q92889 does not annotate explicit “Active site” features at these positions
PTM: Lys-289 N6-acetyllysine added from UniProt Q92889 (missing from original); Ser-521 and Ser-764 phosphoserine added; key-ptms frontmatter updated
Nuclease domain residue range: “658–763” → “658–813 (region); 683–763 (ERCC4 domain core)” per UniProt
Niedernhofer 2006 footnote enriched: F1 hybrid mouse background; Spearman r for transcriptome overlap (r=0.32); precise p-values for IGF-1, glucose, insulin; XFE patient mutation R153P
Vermeij 2016: DR ramp corrected (10% from wk 7 → 30% from wk 9); second-facility ~180% replication added; neocortical neuron retention (~50% more); mouse background (F1 C57BL6J/FVB)
Garinis 2009 footnote enriched: cell types, UV doses, IGF-1R/GHR suppression timing; PDF status updated to “locally available”
Birkisdottir 2021 footnote enriched: all three doses failed (4.7/14/42 ppm); genetic mTOR manipulation also ineffective; p=0.71 for primary arm; PDF status updated

Unverifiable claims:

Deng 2017 (doi:10.1016/j.jflm.2017.05.005) — not in archive; DOI-title match unverified; inherited from ercc1 page; gap-tagged for next lint pass
Open Targets druggability tier 3 — not independently verified via API call

Downstream propagation needed:

molecules/proteins/ercc1.md — cross-check Vermeij 2016 DR ramp protocol and second-facility ~180% replication figure for consistency; ercc1.md holds primary quantitative claims for both Niedernhofer 2006 and Vermeij 2016
hallmarks/genomic-instability.md — if Niedernhofer 2006 cited directly, propagate F1 hybrid background and precise somatotrophic p-values
pathways/nucleotide-excision-repair.md — verify 5’ incision position claim is consistent with Sijbers 1996 stem-loop substrate data

[2026-05-07] verify — molecules/proteins/msh2.md

Pages verified: 1

molecules/proteins/msh2.md — 4 corrections; Lamers 2000 verified against full PDF; Le 2015/Le 2017/Lynch 2009 cross-checked against verified mismatch-repair (no re-read); Fishel 1993 and de Wind 1995 unverifiable

Sources checked:

doi:10.1038/35037523 (Lamers 2000, Nature) — local PDF verified (full 7 pages)
doi:10.1056/NEJMoa1500596 (Le 2015) — cross-checked against verified mismatch-repair page; no re-read
doi:10.1126/science.aan6733 (Le 2017) — closed-access; cross-checked vs mismatch-repair; no-fulltext-access
doi:10.1111/j.1399-0004.2009.01230.x (Lynch 2009) — pending download; cross-checked vs mismatch-repair; pending-verification caveat retained
doi:10.1016/0092-8674(93)90546-3 (Fishel 1993) — closed-access (not_oa); no-fulltext-access
doi:10.1016/0092-8674(95)90319-4 (de Wind 1995) — bronze OA declared but download failed (no URL resolved); no-fulltext-access
UniProt P43246 — confirmed via REST API (934 aa, PTMs, EXO1 region 601–671)
GenAge human — confirmed null (no MSH2 entry) via API
Open Targets ENSG00000095002 — druggability tier 4 confirmed (no approved drug targeting MSH2 directly)

Corrections:

Domain scheme: 5-domain scheme (MBD, Connector, Lever, Clamp, ATPase/dimerization) → 6-region scheme matching Lamers 2000 Fig. 2: MBD (2–115), Connector (116–266), Core+lever helices (267–443 + 504–567), Clamp (444–503), ATPase (568–765), HTH (766–800). “Lever domain” as standalone removed — levers are α-helical elements within the core domain per the paper.
ATPase domain start residue: “~700” → 568 (E. coli MutS; boundary stated explicitly in Lamers 2000 Fig. 2)
Dimerization interface: wiki said “primary heterodimerization interface” is the ATPase domain alone → corrected to two regions: (1) clamp domains and (2) ATPase + HTH; clamp contribution (451 Å² contact surface) now documented.
Lamers 2000 scope caveat added: E. coli MutS structure; domain residue numbers are bacterial, inferences about human MSH2 are by homology mapping — not previously stated on the page.
⚠️ auto-extraction banner removed.

Unverifiable claims:

Fishel 1993 (MSH2 cloning, HNPCC kindred mutations): closed-access; seeder-provided numbers (n HNPCC kindreds, specific mutation details) cannot be confirmed
de Wind 1995 (Msh2-/- lifespan, tumor spectrum): download failed; specific lifespan shortening quantitative claims cannot be confirmed; gap-tagged
Le 2017 53% ORR: closed-access; consistent with mismatch-repair verified page but not independently re-read

Downstream propagation needed:

None required for the structural corrections (domain architecture detail is self-contained on this page)
If any entity page (e.g., a future MSH6 or MSH3 page) cites Lamers 2000 for domain boundaries, the corrected scheme should be applied there too

[2026-05-07] verify — molecules/proteins/lig1.md

Pages verified: 1

molecules/proteins/lig1.md — 10 corrections; 4 sources verified via local PDF; 4 sources closed-access (tagged not_oa)

Sources checked:

doi:10.1038/nature03082 (Pascal 2004, Nature) — local PDF verified
doi:10.1074/jbc.274.47.33696 (Pascucci 1999, JBC) — downloaded + verified (OA)
doi:10.1016/s0960-9822(00)00619-9 (Levin 2000, Curr Biol) — downloaded + verified (OA)
doi:10.1128/mcb.01730-08 (Soza 2009, MCB) — downloaded + verified (PMC OA)
UniProt P18858 — REST API verified
doi:10.1016/0092-8674(92)90450-q (Barnes 1992, Cell) — not_oa
doi:10.1038/ng0896-489 (Bentley 1996, Nat Genet) — not_oa
doi:10.1242/jcs.115.7.1551 (Bentley 2002, JCS) — not_oa
doi:10.1128/mcb.14.1.310-317.1994 (Prigent 1994, MCB) — not_oa

Corrections:

Domain boundaries corrected: N-terminal regulatory domain 1–262 → 1–232 (uncrystallized); DBD 263–535 → 262–535; AdD renamed from “NTase”; domain names aligned to Pascal 2004 Fig. 1b
Crystal structure resolution: “2.2 Å” → “3 Å” (paper states X-ray data refined to 3 Å; Rcryst 23.6%, Rfree 26.8%)
PIP-box residues: “~3–9” → “2–9”; critical residues F8 and F9 (phenylalanines) identified per Levin 2000
Long-patch BER patch size: “2–13 nt” → “2–10 nt maximum; ~70% in 2–4 nt” per Pascucci 1999
Disease variants: “K566E” → “E566K (Glu566Lys)” — the mutation changes Glu to Lys, not Lys to Glu; corrected to match Pascal 2004 and UniProt
Soza 2009 model: “siRNA knockdown” → “46BR.1G1 patient-derived cell line (3–5% of normal LIG1 activity)”; Soza 2009 does not use siRNA
Soza 2009 footnote completely rewritten with correct model, findings (ATM/Chk2 not ATR/Chk1; behind-fork damage; cell morphology)
LIG1 age-related decline claim: attribution to [^soza2009] removed — Soza 2009 does not study aging; claim tagged unsourced
PTM list corrected against UniProt REST API: added Thr233, Ser801, Ser819, Ser911, Ser913 (5 sites missing from seeded list)
Pascal 2004 footnote: “2.2 Å” → “3 Å”; added PDB accession (1X9N); clarified crystallized fragment (residues 233–919) and AppDNA intermediate

Downstream propagation needed:

pathways/base-excision-repair.md table lists “lig1 (LIG1) in nucleus” under short-patch BER key ligase — this is inconsistent with Levin 2000 and Pascucci 1999, which establish LIG1 is specific to long-patch BER; short-patch is LIG3–XRCC1. Main agent should verify and correct BER page table.

[2026-05-07] verify — molecules/proteins/brca2.md

Pages verified: 1

molecules/proteins/brca2.md — 5 factual corrections; 2 closed-access sources tagged no-fulltext-access; 5 local PDFs verified

Sources checked:

doi:10.1038/378789a0 (Wooster 1995, Nature) — local PDF verified
doi:10.1038/386804a0 (Sharan 1997, Nature) — local PDF verified
doi:10.1038/nature01230 (Pellegrini 2002, Nature) — local PDF verified; Pellegrini 2002 substitution for seeder-brief “Yang 2002 Science” confirmed correct
doi:10.1038/nature09399 (Jensen 2010, Nature) — local PDF verified
doi:10.1038/nature03443 (Bryant 2005, Nature) — local PDF verified
doi:10.1126/science.1073834 (Howlett 2002, Science) — not_oa; no-fulltext-access retained
doi:10.1038/nrc.2015.21 (Lord & Ashworth 2016, NRC) — not_oa; no-fulltext-access retained

Corrections:

Chromosome location: “13q12.3” → “13q12-q13 (Wooster 1995); current annotation 13q12.3 (UniProt)” — Wooster 1995 reported 13q12-q13; 13q12.3 is post-1995 refinement
Sharan 1997 lethality timing: “E7.5–E8.5” → “after E6.5; mutant phenotype confirmed at E7.5 (n=45 embryos)” — E8.5 not in Sharan 1997; paper says arrest after day 6.5
Jensen 2010 RPA step: “BRCA2 displaces RPA” → BRCA2 does not directly interact with RPA (pull-down negative per Jensen 2010); indirect displacement is the net outcome; corrected in mechanism step 3
Bryant 2005 xenograft n: “n=40 CD-1 nude mice, 10/group” → 40 implanted (20 per cell line); treatment groups variable n 2–5 per arm per Table 1 (not a clean 10/group)
Sharan 1997 footnote: updated with n=45 embryos, precise strain (AB2.2 ES cells), corrected timing

Unverifiable claims:

Cancer risk percentages (breast ~45–69%, ovarian ~11–17%, etc.) — Antoniou 2003 + Chen & Parmigiani 2007 not in archive; gap-tagged no-fulltext-access in body
Howlett 2002 (FANCD1=BRCA2 identity) — closed-access; claims consistent with known literature but unverified
Lord & Ashworth 2016 BRCAness review — closed-access

Downstream propagation needed:

molecules/proteins/brca1.md — BRCA1-PALB2-BRCA2 axis claims; cross-check that the RPA-displacement mechanism language there is consistent with Jensen 2010 finding (BRCA2 does not directly interact with RPA)
molecules/compounds/ (PARP inhibitor pages, R32b batch) — Bryant 2005 xenograft n correction should propagate to any footnote citing n=10/group for xenograft arm
studies/sharan-1997-brca2-embryonic-lethal stub (if exists) — update timing to post-E6.5, confirmed E7.5

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

Pages verified: 1

molecules/proteins/polb.md — 9 corrections; 3 closed-access sources tagged; 1 source verified via PMC

Sources checked: Sobol 1996 (local PDF), Cabelof 2006 (downloaded), Beard & Wilson 2003 (downloaded), Baptiste 2021 (PMC7887074). Krishna 2005, Patterson/Cabelof 2012, Tomkinson 2001 all not_oa.

Corrections: (1) genage-id null→236; (2) distributive→processive in gapped BER substrates (Sobol 1996 explicit); (3) Beard & Wilson 2003 misattribution for dRP lyase Schiff-base chemistry removed; (4) fidelity 10^-4 misattribution removed — Beard & Wilson 2003 gives kpol not misincorporation rate; (5) Sobol 1996 footnote corrected — 50% BER capacity figure is from Cabelof 2006, not Sobol; (6) Cabelof 2006 claim enriched with n, sex, strain, adenocarcinoma fold, multiple-tumor rate; (7) Baptiste 2021 corrected — “POLB-depleted cells” → Polb-/- MEFs + purified proteins; (8) Sobol 1996 somatic-variant misattribution removed; (9) druggability section updated with Open Targets tractability finding.

Downstream propagation needed: pathways/base-excision-repair.md line 105 (“mitochondrial BER uses Polγ not Polβ”) contradicts Baptiste 2021 — POLB outperforms POLG in mitochondrial gap-filling. Main agent should update BER pathway page.

[2026-05-07] verify — molecules/proteins/ape1.md

Pages verified: 1

molecules/proteins/ape1.md — corrections: 6 (see below); 2 sources closed-access; 2 sources download-failed

Sources checked:

doi:10.1093/nar/gky326 (Li & Wilson 2018, NAR) — downloaded (gold OA via PMC S3) + verified
doi:10.1073/pnas.1000030107 (Stetler 2010, PNAS) — downloaded (green OA via Europe PMC) + verified
doi:10.18632/aging.100810 (Maynard 2015, Aging Albany) — downloaded (hybrid OA) + verified
doi:10.1016/j.jmb.2011.10.023 (Kim 2011, JMB) — downloaded (green OA via IU ScholarWorks) + verified
UniProt P27695 — verified via REST API (length, active-site residues, PTMs)
GenAge HAGRID 195 = APEX1 — confirmed via genomics.senescence.info search
Open Targets ENSG00000100823 — tractability confirmed: SM Advanced Clinical = true; tier 3 confirmed
doi:10.1002/j.1460-2075.1992.tb05411.x (Xanthoudakis 1992, EMBO J) — not_oa; no-fulltext-access tagged
doi:10.1146/annurev.bi.63.070194.004411 (Demple 1994, ARB) — not_oa; no-fulltext-access tagged
doi:10.1093/nar/19.20.5519 (Robson 1991, NAR) — green OA but PDF download failed (no URL resolved); no-fulltext-access tagged
doi:10.1089/ars.2008.2194 (Tell 2009, ARS) — green OA but PDF download failed (no URL resolved); no-fulltext-access tagged

Corrections made:

Mouse model: “APE1 hypomorphic mice (~10% normal APE1 protein in a tissue-conditional context)” → “tamoxifen-inducible conditional KO model (Apex1^flox/flox^Cre-ER+)”. The Li 2018 paper uses conditional KO, NOT hypomorphic mice; “10% protein” phrase is not in the paper.
Mouse model details added: two tamoxifen schedules (postnatal day 7/12, n=4/group; post-weaning week 6, n=6/group); 84% Apex1^-/- lethality before Day 28 in early KO; phenotypic details by schedule.
Maynard 2015 AD APE1 assay n: “53 AD patients vs 30 controls (adjusted for sex and age)” → clarified that the enrolled n was 53/30 but the APE1 incision assay was performed on n=18/group (Table 2 in paper). p-values added (P=0.035 unadjusted; P=0.003 gender-adjusted; P=0.006 gender+age adjusted; P=0.058 age-only adjustment, ns).
Stetler 2010 expanded: PACAP dose (0.2 nmol ICV), signaling pathway (CREB/ATF2 via PKA and p38), n per group (n=8–9 neuronal survival; n=5–6 AP sites), repair-inactive mutant identity (D210A) confirmed.
Active-site residues qualified against UniProt: Y171, D210, H309 are designated active-site; N212 and D308 are binding-site per UniProt; D283 is not in UniProt active/binding site annotation — gap tagged for crystallography literature check.
Kim 2011 mechanism quantified: Km increases 4.4-fold (impaired DNA binding); modification reversible by DTT; in-cell confirmation in HeLa with H₂O₂ 100 µM.

Unverifiable claims:

REF-1 Fos/Jun cysteine assignments (Xanthoudakis 1992; closed-access) — widely accepted, not independently verified
~95% AP-site incision share (Demple 1994; closed-access) — widely accepted, not independently verified
Embryonic-lethal KO and subcellular localization dynamics (Tell 2009; download failed)
APEX1 cDNA cloning details (Robson 1991; download failed)

Downstream propagation needed:

pathways/base-excision-repair.md — if APE1 is described as acting in a “hypomorphic” mouse model, correct to conditional KO; check whether Maynard 2015 n is cited
hallmarks/genomic-instability.md — if it cites APE1 deficiency → aging phenotype, check mouse model description

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

Pages verified: 1

molecules/proteins/pcna.md — corrections: 8 (see below)

Sources checked:

doi:10.1073/pnas.1706508114 (Pilzecker 2017, PNAS) — downloaded (bronze OA) + verified
doi:10.1093/nar/gkaf498 (Rall-Scharpf 2025, NAR) — downloaded (gold OA via PMC) + verified
doi:10.1093/pnasnexus/pgae242 (Shah 2024, PNAS Nexus) — downloaded (gold OA via PMC) + verified
doi:10.1016/j.molcel.2016.12.020 (Choe 2017, Mol Cell) — downloaded (bronze OA) + verified
UniProt P12004 — verified via REST API (length, PTMs, disease variants, cross-references)
Open Targets ENSG00000132646 — queried; no tractability assessments returned (consistent with tier 4)
doi:10.1016/0092-8674(94)90014-0 (Krishna 1994) — not_oa; no-fulltext-access tagged
doi:10.1038/nature00991 (Hoege 2002) — not_oa; no-fulltext-access tagged
doi:10.1038/nrm3562 (Mailand 2013) — not_oa; no-fulltext-access tagged

Corrections made:

ATLD2 variant A252V → removed. UniProt P12004 only records S228I as the ATLD2-associated natural variant. A252V is not in UniProt; wiki claim was unsourced. Gap note added.
K164 SUMO-1 → K164 SUMO2. UniProt records the human K164 site as SUMO2 (not SUMO-1). Hoege 2002 characterized SUMO-1 in yeast; the isoform differs in human. Qualifier added distinguishing yeast (Hoege 2002) from human (UniProt).
“accumulate greater base-substitution mutations” (Pilzecker 2017) → corrected to “functional/cellular evidence of replication-stress-induced genomic instability.” Pilzecker 2017 does not directly quantify mutation rates; it shows γH2AX, cisplatin sensitivity, and HSC cellularity defects. The mutation-accumulation inference is mechanistic extrapolation not directly demonstrated.
Pilzecker 2017 HSC section: added quantitative data from paper — LSK 2.1-fold reduction (41×10³ vs 19×10³ cells/femur), LT-HSC 1.4-fold, ST-HSC 5.3-fold, MPP4 4.4-fold reductions; MPP2 2.1-fold increase (myeloid skew); competitive reconstitution: <5% at 5 weeks vs ~50-60% WT; cisplatin dose 0.8 mg/kg (7.5x below tolerable dose) near-depletes LSK/HSPC; phenotype progresses at 9-10 months.
Mouse strain added: C57BL/6J (stated in Pilzecker 2017 Methods).
Rall-Scharpf 2025 summary expanded with precise cohort n’s, PBL-specific cell type clarification, PCNA western blot sample sizes, and exact IC50 values (POLη: 4.5-fold decrease in old women; PARP inhibitor IQD: 3.3-fold decrease in old men, p<0.0001).
Shah 2024 footnote study design corrected: “in-vitro + in-vivo (mouse + cell lines)” → “in-vivo (mouse genetics: Pcna^K164R intercrosses) + ex-vivo (Trp53kd immortalized MEF cultures).”
Extrapolation table updated: “Phenotype conserved in humans?” changed from “unknown” to “partial” (Rall-Scharpf 2025 PBL poly-Ub reduction + ATLD2 S228I progeroid features provide indirect human support).

Unverifiable claims (closed-access sources):

Krishna 1994: 2.3 Å crystal structure; ~35 Å inner channel diameter; trimeric ring topology. Claims are consistent with downstream review literature (Choe 2017 cites the structure throughout) but primary crystal structure data not directly verified.
Hoege 2002: first demonstration of K164 monoubiquitination, K63-polyubiquitination, and SUMO-1 modification in yeast + mammalian cells. The existence of these modifications is not in doubt (5000+ citations, corroborated by Choe 2017 and UniProt); isoform distinction (SUMO-1 yeast vs SUMO2 human K164) documented.
Mailand 2013: IDCL as PIP-box binding hub; BER/MMR co-localization claims. These are consistent with Choe 2017 review coverage.

Downstream propagation needed:

pathways/mismatch-repair.md — if it describes PCNA’s role and cites SUMO-1, may need SUMO2 correction
cell-types/hematopoietic-stem-cells.md — if it cites Pilzecker 2017 for HSC exhaustion claims, check whether quantitative HSC numbers need updating

[2026-05-07] verify — molecules/proteins/xrcc1.md

Pages verified: 1

molecules/proteins/xrcc1.md — corrections: 5 (see below)

Sources checked:

doi:10.1038/nrg2380 (Caldecott 2008, NRG) — local PDF verified
doi:10.1128/mcb.18.6.3563 (Masson 1998, MCB) — OA PDF downloaded and verified (PMC108937)
doi:10.1038/nature20790 (Hoch 2017, Nature) — OA PDF downloaded and verified (PMC5218588)
doi:10.1128/mcb.10.12.6160-6171.1990 (Thompson 1990) — closed-access, not_oa; retained with no-fulltext-access
doi:10.1093/nar/23.23.4836 (Caldecott 1995) — closed-access, not_oa; retained with no-fulltext-access
doi:10.1016/j.dnarep.2003.08.007 (Tebbs 2003) — closed-access, not_oa; seeder’s correction (no accelerated aging in hypomorphs) corroborated by Hoch 2017 and Masson 1998 cross-references

Corrections made:

NTD boundary “1–151” → “1–183 approx.; Polβ-binding core 84–183” (Masson 1998 Fig. 1A explicitly maps Polβ-binding region to residues 84–183, not 1–151)
BRCT1 mechanism “binds PAR chains” → “direct protein–protein contact with auto-poly(ADP-ribosyl)ated PARP1; preferential for oligo-ADP-ribosylated form” (Masson 1998 primary finding; Caldecott 2008 corroborates — XRCC1 is not a simple PAR-reader)
BRCT1 functional residues: added that deletion analysis maps interaction to 301–402, distinct from UniProt structural annotation 315–403
Hoch 2017 SCAR26 description expanded with specific quantitative data: LIG3α >80% reduced in patient cells; Parp1 deletion rescues interneuron density ~4-fold; rotarod >30-fold improvement to ~30 sec; clarified that NAD+ depletion is the proposed downstream mechanism (elevated ADP-ribose is the measured proxy; direct NAD+ not measured in the paper)
NAD+ mechanism qualified: wiki said “leading to catastrophic NAD+ depletion, bioenergetic failure, and neuronal death” presenting NAD+ depletion as directly demonstrated; corrected to distinguish the paper’s actual measurement (ADP-ribose elevation, PARP1 hyperactivation, neuronal loss rescued by Parp1 deletion) from the proposed mechanism (NAD+ depletion, cited from refs 15–16 in the paper)

Unverifiable claims (closed-access sources):

Thompson 1990: EM9 functional complementation details (radiation sensitivity quantification, SCE levels) — corroborated by Caldecott 2008 review but primary data unverified
Caldecott 1995: LIG3 co-IP and BRCT2 interaction characterization — BRCT2 residue boundary 538–633 corroborated by Masson 1998 Fig. 1A and domain map
Tebbs 2003: full phenotypic data for hypomorphic mice — null lethality (<E7.5) corroborated by Hoch 2017; hypomorph viability corroborated by Masson 1998 ref 49

Downstream propagation needed:

pathways/base-excision-repair.md — XRCC1 scaffold description on that page is consistent with corrections; no propagation required
molecules/proteins/parp1.md — if this page exists and cites XRCC1-PARP1 interaction as PAR-chain binding, may need the same mechanism correction

[2026-05-07] R31b-β — Histone modification + sirtuin family + HDAC inhibitors (DONE; 5/5 pages seeded + verified)

R31b sub-batch β status: COMPLETE. Closes the histone-modification + remaining sirtuin family + HDAC-inhibitor class queue from R31b-α carry-forward.

Pages seeded + verified (5):

processes/histone-modification.md — umbrella process for histone PTMs (acetylation, methylation, phosphorylation, ubiquitination, ADP-ribosylation, citrullination, sumoylation; histone variants H3.3/H2A.Z/macroH2A/H2A.X)
processes/histone-acetylation.md — specific deep-dive on acetyl-Lys marks, HAT/HDAC dynamics, H4K16ac/Sir2/Sas2 yeast lifespan axis, acetyl-CoA metabolic-epigenetic bridge
molecules/proteins/sirt2.md — primarily cytoplasmic sirtuin; α-tubulin K40 deacetylase; M-phase H4K16ac eraser; CDH1+CDC20 (APC/C activator) substrates; AGK2-style inhibition direction for neurodegeneration (distinct from SIRT1/6 activator pattern)
molecules/proteins/sirt7.md — nucleolar sirtuin; primary activity is H3K122 desuccinylation (Li 2016, NOT H3K18 as my brief asserted); secondary deacetylation; SIRT7-NRF1-mtUPR-HSC axis (Mohrin 2015) is the load-bearing aging mechanism
interventions/pharmacological/hdac-inhibitors.md — class page (vorinostat, romidepsin, belinostat, panobinostat, chidamide; HDAC6-selective ricolinostat/tubastatin; aliphatic acids butyrate/valproate); human-evidence-level: none for aging indication — all FDA approvals are oncology

R31b-β verification statistics:

5 pages verified end-to-end; ~30 source-text-level corrections; ~12 scientifically critical, including:
- histone-acetylation (most consequential): H4K16ac aging direction REVERSED in wiki — actually INCREASES at subtelomeres in old yeast cells (was “declines”); sas2Δ + SIR2-OE both EXTEND lifespan (was inverted); H4K16Q mimics permanent acetylation and SHORTENS lifespan; Das 2009 cell-cycle framing corrected (H3K56ac present throughout cell cycle, not exclusively damage-context); SIRT2 added as primary H3K56ac eraser (not SIRT1); Feser 2010 mechanism corrected (histone protein stoichiometry, NOT H3K56ac restoration); lifespan numbers (~30-50% extension WT, 65% asf1 rescue).
- histone-modification: druggability tier 1→2 for consistency with hdac.md; SIRT1/H4K16ac substrate confused with yeast Sir2 (mammalian SIRT1 is primarily H3K9ac per Pal & Tyler 2016); Feser 2010 lifespan figures corrected; Michishita 2008 SIRT6/H3K9ac telomere claims qualified.
- SIRT2: substrate BubR1 → CDH1 + CDC20 (Kim 2011); catalytic-dead mutant S368A → N168A/H187Y (North 2003); CDK1 → CDK2/CDK5 for S368 phosphorylation; cell lines corrected (no COS-7, no U2OS); PR-Set7 K90 site specified (Serrano 2013); strain-discordance documented (Kim 2011 mixed-lineage tumors vs Serrano 2013 C57BL/6 no-tumors).
- SIRT7 (most consequential): H3K18 → H3K122 succinylation (my brief had wrong substrate position; Li 2016 actually shows H3K122-succinyl at DSB sites recruited via PARP1); SIRT7-NRF1 directional inversion: NRF1 recruits SIRT7 (NOT vice versa as wiki said) — NRF1 is the targeting TF, SIRT7 the co-recruited transcriptional repressor at mRP/mTF promoters; Vakhrusheva 2008 specific lifespan numbers (~12 mo vs ~24 mo) removed (not in abstract; closed-access paper); Mohrin 2015 transplantation metric added (~40% reconstitution reduction, n=7); Vazquez 2019 PNAS HGPS-rescue claim dropped (paper does not exist; closest match is Sun 2020 Sci Adv with different authors/year).
- hdac-inhibitors: EPIC trial (NCT07404085) drug identity was unspecified — actual is sodium butyrate (specific, not “an HDAC inhibitor”); population is remitted mood disorder patients aged 40–75 (n≈160); Zhang 2014 11-author list compressed to 2 authors → corrected; clinical-trials-active=1 confirmed via direct ClinicalTrials.gov v2 query.
~6 wrong DOIs/citations caught: Vakhrusheva 2008 (CIRCRESAHA.108.170399 → 107.164558), Mohrin 2015 (science.aab2307 → aaa2361), Ford 2006 (gad.1364706 → 1399706), Li 2016 journal (Nature → Nature Communications); Dang 2009 vs Das 2009 DOI confusion (10.1038/nature07861 is Das 2009; Dang 2009 is 10.1038/nature08085 — both papers confirmed via Crossref); “Mardalsa & Wellen 2017” not in PubMed → substituted with Sivanand & Wellen 2018 TIBS; “Berger 2007 Nature” + “Sun 2014 review” both unconfirmable → omitted.

Cross-cutting propagation done by main agent (R31b-β post-verification pass):

processes/histone-acetylation.md — Dang 2009 footnote text corrected: “declines” → “increases” (verifier fixed body but missed footnote summary)
molecules/proteins/sirt1.md — § Histones updated to clarify mammalian SIRT1 is primarily H3K9ac (Pal & Tyler 2016); H4K16ac is secondary in mammals and primary in yeast Sir2
pathways/sirtuin.md — family substrate table updated: SIRT1 H3K9ac primary / H4K16ac secondary-mammals-primary-yeast; SIRT2 expanded with CDH1+CDC20+PR-Set7-K90 substrates; SIRT7 H3K122-succinyl primary
hallmarks/epigenetic-alterations.md — § “SIRT1 decline and histone deacetylation loss” updated with mammalian H3K9ac primary / H4K16ac context-dependent clarification

SOP-effectiveness check: The new wiki-verifier SOP § 3 was applied effectively across the batch:

hdac-inhibitors verifier: explicitly cross-checked McIntyre 2019, McDonald 2013, Pao 2020 against verified [[hdac]] page (R31) without re-reading the PDFs (“All claims consistent. No PDF re-read per SOP § 3.”). This is a clear demonstration of the new SOP saving PDF reads.
histone-modification verifier: leveraged verified [[sirt1]] and [[sirt6]] pages for Vaquero 2007 / Michishita 2008 SIRT6 claims to reduce duplicate verification work
histone-acetylation verifier: cross-checked SIRT1/SIRT6 substrate claims against verified family pages

Schema escalations (carry-forward):

Druggability-tier convention (R26-R27) is now applied consistently: process pages should not exceed their primary-target protein page’s tier. histone-modification and histone-acetylation both pulled tier 1 → 2 to match [[hdac]] (R31 tier 2).
The seeder-brief DOI memory unreliability pattern continues (R29-R31b-β = ~50 wrong DOIs caught across these rounds); worth elevating to a structural CLAUDE.md note.

Hallmark-level impact (cumulative R31 + R31b):

epigenetic-alterations mechanistic depth before R31: 12 prot / 2 path / 1 proc / 7 interv
After R31 + R31b-α + R31b-β: ~16 prot / 2 path / 5 proc / 8 interv. The methylation/demethylation/histone-modification/histone-acetylation process layer is now seeded; HDAC-inhibitor class added to interventions (gives the hallmark a +1 intervention class).

Remaining gaps in this cluster (deferred):

HDAC family member individual pages (HDAC1–11) — class page suffices unless aging-specific primary literature accumulates per member
TDG protein page — referenced from dna-demethylation.md as required for OSK rejuvenation per Lu 2020; SIRT4, SIRT5 — sirtuin family completion (SIRT1, 2, 3, 6, 7 done; 4 + 5 are mitochondrial metabolic-regulation focused, lower aging-leverage)

[2026-05-07] verify — molecules/proteins/sirt7.md

Page verified: molecules/proteins/sirt7.md

Sources checked:

Ford 2006 doi:10.1101/gad.1399706 — downloaded via Europe PMC (PMC1472467); full PDF read. OA (diamond).
Li 2016 doi:10.1038/ncomms12235 — downloaded via PMC S3 (PMC4961794); full PDF read. OA (gold).
Mohrin 2015 doi:10.1126/science.aaa2361 — closed-access (not_oa in archive); full text verified via PMC4447312.
Vakhrusheva 2008 doi:10.1161/CIRCRESAHA.107.164558 — closed-access (not_oa); verified against PubMed abstract PMID 18239138 + Crossref metadata.
Barber 2012 doi:10.1038/nature11043 — closed-access (not_oa); verified against PubMed abstract PMID 22722849 + Crossref metadata.
UniProt Q9NRC8 — confirmed via REST API (400 aa, deacylase domain 82-329).
GenAge ID 269 — confirmed via direct query.
Open Targets ENSG00000187531 — tier 4 confirmed (no approved drug, no advanced clinical, no Phase 1 clinical, no high-quality ligand).

Corrections made:

CRITICAL: H3K18 desuccinylation → H3K122 desuccinylation throughout, for all Li 2016-derived claims. Li 2016 characterizes SIRT7 as an H3K122 desuccinylase at DSB sites (PARP1-dependent); H3K18 deacetylation belongs to Barber 2012. These are different residues, different marks, different cellular functions. The error appeared in: (1) opening paragraph, (2) Enzymatic activities §1, (3) Genome stability aging section, (4) cross-references/epigenetic-alterations, (5) Li 2016 footnote.
Ford 2006 enzymatic mechanism: added explicit statement that recombinant SIRT7 showed NO detectable deacetylase or ADP-ribosyltransferase activity in vitro (directly from the paper text) — the Pol I activation mechanism was unresolved in 2006.
Mohrin 2015 mechanism direction: corrected “SIRT7 activates NRF1” → “NRF1 recruits SIRT7 to mRP/mTF gene promoters” — the directionality was inverted. NRF1 is the targeting factor, SIRT7 is the transcriptional repressor co-recruited by NRF1.
Mohrin 2015 transplantation data: added specific “~40% reduction in long-term hematopoietic reconstitution (n=7)” from PMC full text.
Vakhrusheva 2008 lifespan numbers: removed unverifiable “~12 months vs ~24 months WT” — abstract only states “reduction in mean and maximum lifespans” without specific values. Added gap tag needs-replication.
Vakhrusheva 2008 cardiomyocyte apoptosis: added “~200% increase in basal apoptosis” from abstract (stated as “approximately 200%” in source), and 3-fold collagen III increase.
Footnotes: added PMID and PMC IDs for Mohrin 2015 (PMID:25792330, PMC:PMC4447312) and Barber 2012 (PMID:22722849, PMC:PMC3412143) and Vakhrusheva 2008 (PMID:18239138).
Ribosome biogenesis section: removed unsupported claim that SIRT7 “desuccinylates H3K18 at rDNA promoters” — that specific claim is not from Li 2016 or Ford 2006.
Cross-references: corrected SIRT1/SIRT6 comparative notes to properly distinguish H3K18 deacetylation (Barber 2012) from H3K122 desuccinylation (Li 2016).
cGAS-STING gap: removed incorrect “H3K18 desuccinylation / L1 repression” framing; replaced with honest statement that the causal chain is undemonstrated.

Unverifiable claims:

Vakhrusheva 2008 specific lifespan medians — abstract-only; paper closed-access.
Barber 2012 detailed cell-type breakdown (which cancer lines showed strongest effect) — abstract-only.
Any claim about SIRT7 deacetylating ATM, CDK9, DDX21 as “non-histone substrates” — these are stated in the wiki body but not sourced to any of the 5 primary citations on this page; would need additional citations.

Vazquez 2019 PNAS drop confirmed: No paper matching “Vazquez 2019 PNAS SIRT7 HGPS rescue” found via PubMed, Semantic Scholar, or Crossref. The relevant paper is Sun et al. 2020 Science Advances (doi:10.1126/sciadv.aay5556) — different authors, different journal, different year. Seeder’s decision to drop this citation was correct.

Downstream propagation needed:

molecules/proteins/nrf1.md — if it exists, may describe SIRT7 as downstream of NRF1; check the directional language now confirmed as NRF1→recruits→SIRT7.
processes/unfolded-protein-response.md or mtUPR section — any mention of “SIRT7 activates NRF1” should be corrected.
cell-types/hematopoietic-stem-cells.md — if it cites Mohrin 2015, the 40% reconstitution deficit figure and the NRF1 recruitment mechanism should be cross-checked.
hallmarks/epigenetic-alterations.md — may have inherited the H3K18 desuccinylation error from this page.
studies/li-2016-sirt7-desuccinylase.md — stub study page likely has wrong substrate (H3K18 vs H3K122); should be seeded correctly now that PDF is verified.

Final verified state: verified: true with scope (Ford 2006 and Li 2016 verified via full PDF; Mohrin 2015 verified via PMC full text; Vakhrusheva 2008 and Barber 2012 verified via PubMed abstract + Crossref; Vakhrusheva 2008 specific lifespan numbers unverifiable from abstract).

[2026-05-07] verify — molecules/proteins/sirt2.md

Page verified: molecules/proteins/sirt2.md

Sources checked:

North 2003 doi:10.1016/s1097-2765(03)00038-8 — local PDF downloaded and read (bronze OA); full corrections applied
Vaquero 2006 doi:10.1101/gad.1412706 — local PDF downloaded and read (diamond OA via Europe PMC); full corrections applied
Kim 2011 doi:10.1016/j.ccr.2011.09.004 — local PDF downloaded and read (bronze OA via Cell Press); full corrections applied
Serrano/Vaquero 2013 doi:10.1101/gad.211342.112 — local PDF downloaded and read (diamond OA via UAB repository); full corrections applied
Outeiro 2007 doi:10.1126/science.1143780 — download failed (green OA, 0 candidate URLs); abstract verified via Crossref; tagged no-fulltext-access
Wang 2019 doi:10.1016/j.arr.2019.100961 — closed-access confirmed (not_oa); tagged no-fulltext-access
UniProt Q8IXJ6 — re-verified via REST API (domain boundaries, subcellular locations, PTMs)
Open Targets Platform ENSG00000068903 — queried; druggability tier 2 confirmed appropriate for aging context

Critical corrections:

BubR1 → CDH1 + CDC20 (Kim 2011 APC/C mechanism). Wiki stated “SIRT2 deacetylates BUBR1 and CDH1.” Kim 2011 shows SIRT2 deacetylates CDH1 and CDC20 (the actual APC/C coactivators). BubR1 is not a SIRT2 substrate in this paper. Body text, pathway table, and footnote all corrected.
S368A mutant removed from North 2003 section. Wiki cited “SIRT2-S368A (phosphorylation-dead mutant)” as a North 2003 result. North 2003 uses N168A and H187Y (catalytic dead mutants); S368A is not used or mentioned. Replaced with accurate description of actual North 2003 findings including N168A, H187Y, siRNA knockdown, and enzymatic kinetics.
CDK1 → CDK2 + CDK5 (S368 phosphorylation). Frontmatter listed Ser368-phosphorylation-CDK1 — UniProt Q8IXJ6 records S368 phosphorylation by CDK2 and CDK5 (not CDK1). Also S372 phosphorylation added. Frontmatter corrected.
North 2003 cell lines corrected. Footnote listed “COS-7, HeLa” — North 2003 uses human fibroblasts (GM03318), HeLa, and 293T cells. COS-7 is not used in this paper. Footnote corrected.
Vaquero 2006 cell lines corrected. Footnote listed “HeLa, U2OS” — Vaquero 2006 uses HeLa and 293/293F cells. U2OS not used. Footnote corrected.
Kim 2011 sample sizes added. Tumor survival curves: female n=26 KO, 13 het, 22 WT; male n=19 KO, 10 het, 20 WT. Cancer incidence ~60% by 24 months (female) and ~60% by 20 months (male). Added to footnote and body.
Kim 2011 mixed-lineage strain noted. Kim 2011 spontaneous tumor phenotype occurs in mixed-lineage mice; Serrano 2013 C57-BL6 mice did NOT develop spontaneous tumors. Strain-background discordance documented in body text and gap tags.
Serrano 2013 PR-Set7 mechanism specified. Wiki said “deacetylates PR-Set7”; paper identifies K90 as the specific deacetylation site. Body text and footnote corrected to include K90 and the downstream H4K20me1→me2→me3 cascade mechanism.
Domain boundaries de-approximated. UniProt confirms sirtuin domain is residues 57–338 exactly; “~57–338” approximation qualifier removed.
Outeiro 2007 body text qualified. Cell line name (SK-N-MC) and AGK2 concentration not confirmable from Crossref abstract; added no-fulltext-access caveat in body and footnote.

Unverifiable (permanent):

Wang 2019 (closed-access, not_oa) — SIRT2 age-related expression decline in brain (claimed in wiki) remains unchecked against full review text
Outeiro 2007 (download failed, green OA) — exact cell line names, AGK2 IC50/concentration, specific inclusion morphology quantitative data unchecked

Downstream propagation needed (for main agent):

processes/histone-acetylation.md — already references SIRT2 as H4K16ac eraser; check if CDH1/CDC20 correction or APC/C mechanism needs updating there
studies/kim-2011-sirt2-tumor-suppressor-apc-c.md — if this study page exists, the BubR1 → CDH1/CDC20 correction must be propagated
studies/north-2003-sirt2-tubulin-deacetylase.md — if this study page exists, verify it does not mention S368A mutant as North 2003 result

[2026-05-07] verify — processes/histone-acetylation.md

Page verified: processes/histone-acetylation.md

Sources checked (5 local PDFs read end-to-end):

Dang 2009 doi:10.1038/nature08085 — local PDF; critical corrections (see below)
Imai 2000 doi:10.1038/35001622 — local PDF; NAD+-dependent mechanism + O-acetyl-ADP-ribose product confirmed correct
Das 2009 doi:10.1038/nature07861 — local PDF; CBP/p300 writer confirmed; eraser corrected (SIRT2 primary); cell-cycle framing corrected; corrigendum (Nature 460:1164) noted
Feser 2010 doi:10.1016/j.molcel.2010.08.015 — downloaded + verified; lifespan numbers added (30–50%); footnote corrected from “restores H3K56ac” to histone-stoichiometry mechanism
Pal & Tyler 2016 doi:10.1126/sciadv.1600584 — downloaded + verified; review synthesis spot-checked, consistent

Unverifiable (download failed, green OA): Sivanand 2018 + Sen 2019 — tagged no-fulltext-access

Critical corrections:

H4K16ac aging direction inverted → corrected. Wiki stated “H4K16ac is progressively lost at subtelomeric chromatin.” Dang 2009 Fig. 1c shows H4K16ac increases at X elements in old cells; Sir2 protein decreases. Table corrected accordingly.
H4K16ac lifespan genetics inverted → corrected. Wiki stated sas2 OE increases H4K16ac and sas2Δ decreases it — opposite of paper. Dang 2009: sas2Δ extends lifespan (mean 32.4 vs 26.7 WT); SIR2-OE extends (31.4); H4K16Q shortens (11.7). Entire section rewritten.
Das 2009 cell-cycle framing. H3K56ac is not exclusively a “DNA-damage context” mark — Das 2009 shows it is present throughout the cell cycle; damage increases it.
SIRT2 added as primary H3K56ac eraser per Das 2009 (SIRT2 more active than SIRT1 in vitro).
Feser 2010 mechanism corrected. Not “restores H3K56ac” but histone stoichiometry; pathway distinct from Sir2/ERC and CR.
Druggability-tier 1 → 2 (aging-context convention, consistent with hdac R31).

Downstream propagation needed (for main agent):

hallmarks/epigenetic-alterations.md — check for H4K16ac aging direction claims

[2026-05-07] verify — processes/histone-modification.md

Page verified: processes/histone-modification.md

Sources checked (6 cited):

Michishita 2008, doi:10.1038/nature06736 — locally available; PDF read in full (4 pages). Verified.
Pal & Tyler 2016, doi:10.1126/sciadv.1600584 — gold OA; downloaded and PDF read in full (10 of 19 pages covering all histone-modification sections). Verified.
Feser 2010, doi:10.1016/j.molcel.2010.08.015 — bronze OA; downloaded and PDF read in full (12 pages). Verified.
Allis & Jenuwein 2016, doi:10.1038/nrg.2016.59 — not_oa; permanently unverifiable. Claims tagged no-fulltext-access.
Vaquero 2007, doi:10.1038/sj.onc.1210617 — not_oa; permanently unverifiable. Claims tagged no-fulltext-access.
De Cecco 2019, doi:10.1038/s41586-018-0784-9 — not_oa; permanently unverifiable. Claims tagged no-fulltext-access.

Seeder DOI omissions confirmed correct: “Berger 2007 Nature” and “Sun 2014 review” omitted by seeder per user brief — neither resolves to a real paper on the claimed topic. Omission confirmed appropriate.

Corrections made (6):

druggability-tier: 1 → 2 (aging-context correction, mirrors hdac R31; FDA-approved HDAC inhibitors are oncology-only; no aging-validated drug exists — consistent with CLAUDE.md aging-context convention).
Feser 2010 lifespan extension figure: “~40%” → “30–50% in wild-type yeast depending on galactose induction level” — “~40%” is not stated anywhere in the paper. Actual figures: ~30% at 1% galactose (Fig. 6A, p=0.0071), ~50% at 2% galactose (Fig. 6B, p=0.0068); asf1 mutant rescue with all-4-core-histone plasmid = 65% extension (Fig. 3E). The wiki also incorrectly said “a single extra copy of the H3–H4 gene pair” — the experiments used high-copy plasmids or galactose-inducible promoters, not a single extra copy.
Michishita 2008 mechanism qualifier added — the wiki claimed SIRT6 deacetylation of H3K9ac at telomeres “promotes the conversion to H3K9me3 by downstream KMTs.” The paper does not demonstrate or claim this conversion; it shows SIRT6 maintains low H3K9ac at telomeres required for WRN association and efficient replication. The H3K9me3 downstream inference is from the broader literature, not this paper. Text corrected to reflect what the paper actually shows (SIRT6-mediated H3K9ac deacetylation maintains WRN occupancy and prevents TIF formation; chromosome fusions when SIRT6 is depleted).
SIRT1/H4K16ac primary-substrate attribution corrected — the wiki stated H4K16ac is “the primary histone substrate of SIRT1 in its heterochromatin-maintenance roles.” Pal & Tyler 2016 (the cited source) explicitly states mammalian SIRT1 is primarily an H3K9ac deacetylase; H4K16ac deacetylation is the primary substrate of yeast Sir2 (the SIRT1 ortholog). H4K16ac deacetylation in mammals is associated with SIRT2, not SIRT1. H4K16ac table row updated; explanatory paragraphs corrected to distinguish yeast (Sir2/H4K16ac) from mammalian (SIRT1/H3K9ac) axes.
SIRT1/NAD+ section retitled and corrected — “SIRT1 and H4K16ac — the NAD+ connection” retitled and corrected to distinguish yeast Sir2/H4K16ac from mammalian SIRT1/H3K9ac mechanisms.
Footnotes updated — Michishita 2008 and Feser 2010 footnotes expanded with actual n values, p-values, and experimental details from the PDFs; Pal & Tyler 2016 footnote updated to note the SIRT1/H3K9ac vs Sir2/H4K16ac distinction; download-pending statuses removed.

Unverifiable claims (3 not_oa sources):

Allis & Jenuwein 2016: foundational histone code framing (background/conceptual, low risk of error).
Vaquero 2007: H4K16ac aging claim in mammals via sirtuins — this is the primary citation for the mammalian H4K16ac/SIRT1 axis that was partially corrected based on Pal & Tyler 2016 contradiction; Vaquero 2007 may itself support a broader sirtuin/H4K16ac claim, but the primary source cannot be read. The correction narrows the claim to yeast.
De Cecco 2019: L1/cGAS-STING heterochromatin-loss axis; all claims kept with no-fulltext-access tag.

Downstream propagation needed (main agent):

[[sirt1]] (verified page) — the H4K16ac/H3K9ac substrate clarification is consistent with what the sirt1 page says (it correctly lists both H3K9ac and H4K16ac), but the sirt1 page may benefit from a clarifying note that H4K16ac is the yeast substrate and H3K9ac is the primary mammalian substrate. Low priority given sirt1 is already verified.
[[sirtuin]] — may carry similar H4K16ac/SIRT1 conflation from the yeast→mammal translation. Worth checking.
Any page citing [^feser2010] for “~40% lifespan extension” — check for inherited incorrect figure.

Final state: verified: true with scope — Michishita 2008, Pal & Tyler 2016, Feser 2010 verified against PDFs. Allis & Jenuwein 2016, Vaquero 2007, De Cecco 2019 not_oa and permanently unverifiable; claims from these sources remain tagged no-fulltext-access.

[2026-05-07] verify — interventions/pharmacological/hdac-inhibitors.md

Page verified: interventions/pharmacological/hdac-inhibitors.md

Sources checked:

McIntyre 2019 (doi:10.15252/emmm.201809854) — cross-checked against verified hdac page (R31); no PDF re-read per SOP § 3. All claims consistent.
McDonald 2013 (doi:10.1016/j.exger.2012.09.006) — cross-checked against verified hdac page (R31); no PDF re-read per SOP § 3. Drug identity (vorinostat/SAHA; not sodium butyrate) and fly strains (Canton-S, Oregon-R) confirmed consistent.
Pao 2020 (doi:10.1038/s41467-020-16361-y) — cross-checked against verified hdac page (R31); no PDF re-read per SOP § 3. All mechanistic and quantitative claims consistent (exifone 50 mg/kg/day IP 4 wk; 17-mo WT mice for OGG1/8-oxoG; 8-mo 5XFAD for behavioral rescue). HDAC1-activation nuance correctly rendered on both pages.
Marks 2007 (doi:10.1038/sj.onc.1210204) — not_oa. Vorinostat history, Zn2+-chelation mechanism, and 2006 FDA-approval claims corroborated against verified hdac claims; full text unverifiable. Remains as partial-source.
Zhang et al. 2014 (doi:10.3233/JAD-140066) — not_oa (closed access). Full text unavailable. Author list corrected (seeder had “Zhang L, Qin C” → “Zhang L et al., 11 authors, last/corresponding: Qin C” per Crossref). Paper title confirmed: “Tubastatin A/ACY-1215 Improves Cognition in Alzheimer’s Disease Transgenic Mice.” Quantitative claims (n, strain, doses, outcomes) unverifiable; tagged no-fulltext-access.
NCT07404085 (EPIC trial) — verified RECRUITING via ClinicalTrials.gov v2 API. Drug identified as sodium butyrate (not an unspecified HDAC inhibitor drug); trial uses 40–75-year-old mood disorder patients (not healthy aging adults); estimated n=160; primary endpoint: 3-week global cognitive composite. Page description updated to include these specifics.
clinical-trials-active: 1 — confirmed. Three ClinicalTrials.gov v2 searches run (“HDAC inhibitor aging”, “histone deacetylase inhibitor older adults”, “HDACi longevity”); only NCT07404085 qualifies as aging-adjacent. NCT02787369 (ricolinostat in CLL; ACTIVE_NOT_RECRUITING) confirmed oncology-only and excluded.

Corrections made (4):

Zhang 2014 author attribution: “Zhang L, Qin C” → “Zhang L et al. (11 authors; last/corresponding: Qin C)”. Seeder abbreviated incorrectly, implying a 2-author paper.
Zhang 2014 archive status: “not confirmed” → “not_oa (closed-access; full text not verified)”; added no-fulltext-access tag.
EPIC trial drug identity: body description updated to specify sodium butyrate as the HDAC inhibitor used (not a named clinical HDAC inhibitor drug); age range (40–75), population (mood disorder patients), and estimated n=160 added; note that this is a non-selective aliphatic acid rather than a clinical-grade agent.
clinical-trials-active inline note: added documentation of v2 search queries and confirmation that NCT02787369 was evaluated and excluded (oncology-only).

Consistent with verified sources (no correction needed):

human-evidence-level: none — confirmed correct per schema (no completed aging-indication trials; CLAUDE.md schema defines “limited-negative” for completed Phase 2/3 failures, which the oncology failures are not)
literature-checked-through: 2026-05-07 — consistent with seeder’s claimed PubMed searches
Intervention-classes.md hdac-inhibition block — confirmed well-formed; Dataview query block correct; all constituent compound links consistent

Unverifiable claims:

Zhang 2014: specific n, mouse strain, doses, quantitative cognitive outcomes — not_oa; full text permanently inaccessible
Marks 2007: all claims derived from it — not_oa; corroborated against hdac R31 verified page but primary source unread

Downstream propagation needed (main agent):

None identified. All McIntyre/McDonald/Pao claims are consistent with hdac which is the canonical verified source. No numeric corrections that would propagate. The Zhang 2014 author-list correction is footnote-local.

Final state: verified: true with scope — McIntyre/McDonald/Pao verified via hdac cross-check; Marks 2007 (not_oa) corroborated but primary source unread; Zhang 2014 (not_oa) — author list corrected, quantitative claims tagged no-fulltext-access.

[2026-05-07] R31b-α — DNA methylation cluster (DONE; 5/5 pages seeded + verified)

R31b sub-batch α status: COMPLETE. Closes the DNA methylation cluster identified in the R31 ROADMAP carry-forward.

Pages seeded + verified (5):

studies/schooling-2025-mr-epigenetic-clock-lifespan.md — MR null result for clock-acceleration → mortality (5 inbound; previously cited from information-theory-of-aging.md and partial-reprogramming.md without dedicated page)
processes/dna-methylation.md — central process page covering DNMT3A/B/L de novo, DNMT1/UHRF1 maintenance, CpG island/CpG shore landscape, age-related global hypomethylation + focal hypermethylation, foundational basis for all epigenetic clocks
processes/dna-demethylation.md — TET-TDG-BER active demethylation axis (5mC→5hmC→5fC→5caC→C); 5hmC as stable mark; passive demethylation; aging-relevant TET2/CHIP and OSK-required demethylation
molecules/proteins/dnmt1.md — maintenance methyltransferase; UHRF1 partnership; HSAN1 + ADCA-DN disease mutations; SIRT1 cross-talk
molecules/proteins/nanog.md — homeodomain pluripotency TF; OCT4-SOX2-NANOG core circuit; in 5-factor (Yu 2007) and 6-factor (Lapasset 2011 OSKMNL) cocktails — NOT in OSKM (Ocampo 2016) or OSK (Lu 2020 / Yang 2023 — confirmed R31)

R31b-α verification statistics:

5 pages verified end-to-end; ~25 source-text-level corrections; ~10 scientifically critical, including:
- DNA-methylation: Heyn 2012 magnitude was ~2× too high (“~20%” → ~9% relative; 80.5%→73.0%); sample sizes fabricated (actual WGBS n=1+1, array n=19+19); Bocklandt 2011 model uses 2 CpG sites (not 3).
- DNMT1: HSAN1 mutation “P491L” → p.Asp490Glu-Pro491Tyr (P491L not in Klein 2011); mechanism “weakens HP1 binding” → “abolishes heterochromatin binding after S phase”; narcolepsy belongs to ADCA-DN not HSAN1; Bocklandt 2011 attribution backwards — paper does NOT support stochastic intra-pair drift (intra-pair p>0.1).
- NANOG: OCT4-NANOG co-occupancy “>80%” → 44.5% (345 of 776) per Loh 2006 Fig. 3b; method ChIP-chip → ChIP-PET.
- DNA-demethylation: 5hmC unit ”% of all C” → % of dG; brain values 0.3–0.7% of dG (was 0.4–1.5% of all C); Jaiswal 2014 CHIP prevalence numbers + framing reversed (was TET2-specific 2-3%/5%; actual overall 5.6%/9.5%/11.7%/18.4% at decades 60–90+).
- Schooling 2025: GrimAge instrument count 4–11 → 4 SNPs; r² variance attribution corrected (3% is for telomere length, not clock instruments).
~5 wrong DOIs/years caught (Jaiswal 2014 NEJMoa1409617→1408617; Song 2012→2010; Wilson & Jones 1983 DOI located; Kohli & Zhang 2013 substituted for unconfirmable Bagci & Fisher 2013).

Light propagation: hallmarks/epigenetic-alterations.md Schooling 2025 stub-annotation upgraded; hypotheses/information-theory-of-aging.md verified-scope updated to mark all 5 study pages now exist + verified. P491L grep showed only log.md (corrections record); no other propagation needed. Both verifiers (DNMT1 + DNA-methylation) independently fixed their own pages’ Bocklandt 2011 framings — no further propagation needed.

SOP-effectiveness check: The new wiki-verifier SOP § 3 (study-page-first) was applied successfully — DNA-demethylation verifier explicitly stated “Lu 2020 cross-checked against verified study page; no PDF re-read needed.” NANOG verifier: “Lapasset 2011 cross-checked against verified processes/partial-reprogramming.md; no PDF re-read per SOP.” First round to demonstrate the SOP working as intended.

Schema escalations: study-design: mendelian-randomization not in CLAUDE.md enumerated list; multi-arm n-subjects: representation continues to need formalization.

Hallmark-level impact: epigenetic-alterations 12/2/1/7 → ~14/2/3/7 (cumulative R31+R31b-α). Methylation/demethylation process layer now seeded.

R31b-β remaining (queued): histone-modification + histone-acetylation processes; sirt2 + sirt7 proteins; hdac-inhibitors class page.

[2026-05-07] verify — molecules/proteins/dnmt1.md

Page verified: molecules/proteins/dnmt1.md

Sources checked:

Klein 2011 (doi:10.1038/ng.830) — local PDF verified end-to-end (HSAN1 mutations, mechanism)
Sharif 2007 (doi:10.1038/nature06397) — local PDF verified end-to-end (UHRF1/Np95 SRA mechanism)
Bocklandt 2011 (doi:10.1371/journal.pone.0014821) — local PDF verified end-to-end (methylation-age correlation)
UniProt P26358 — REST API verified (all domain boundaries match exactly)
Open Targets ENSG00000130816 — API queried (Approved Drug SM = true; aging-context tier 2 confirmed)
GenAge — DNMT1 absent confirmed via live query (Build 21, 2023)
Song 2010 (doi:10.1126/science.1195380) — not_oa; PDF unavailable; tagged no-fulltext-access
Bostick 2007 (doi:10.1126/science.1147939) — not_oa; unverifiable
Bestor 1988 (doi:10.1016/0022-2836(88)90122-2) — not_oa; unverifiable
Winkelmann 2012 (doi:10.1093/hmg/dds035) — PDF pending; unverified

Corrections made:

HSAN1 mutation “P491L” → “p.Asp490Glu-Pro491Tyr” (triple nucleotide change c.1470TCC-1472ATA causing two aa changes on one allele; P491L does not appear in Klein 2011)
HSAN1 mechanism “weakens HP1 binding” → “abolishes heterochromatin binding after S phase” (Klein 2011 describes heterochromatin-binding loss, not HP1 binding specifically)
Removed narcolepsy from HSAN1 clinical features (narcolepsy is an ADCA-DN feature per Winkelmann 2012, not reported in Klein 2011 HSAN1 Table 1)
Added quantitative clinical onset data from Klein 2011 Table 1 (symptom onset 16–35, dementia 30s–40s)
Added 8% global hypomethylation quantification (p<0.001) from Klein 2011
UHRF1 step 2 mechanism: removed unsourced “tandem Tudor and PHD finger + ubiquitinated H3” detail (not in Sharif 2007; tagged gap)
Bocklandt 2011 “epigenetic drift” attribution: corrected — Bocklandt explicitly failed to find intra-twin-pair drift (p>0.1); changed to “directional locus-specific methylation changes”
Bocklandt footnote: corrected n to “34 pairs (68 individuals) discovery + 60 replication”; added key finding that drift was NOT replicated
Song 2010 footnote: updated to reflect not_oa status and tagged no-fulltext-access
UniProt footnote: updated to note domain boundaries re-verified against live API
Druggability tier: confirmed correct via Open Targets API; removed gap tag, added confirmation note
GenAge null: confirmed via live query; removed needs-canonical-id for GenAge

Downstream pages potentially affected:

processes/dna-methylation.md — may cite Bocklandt 2011 for drift claims; check for same misattribution
hallmarks/epigenetic-alterations.md — may inherit Bocklandt drift claim
molecules/biomarkers/horvath-clock-2013.md — cites methylation-drift concept; no correction propagation needed (Bocklandt correction is about mechanism, not clock)

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

Page verified: molecules/proteins/nanog.md

Sources checked:

Chambers 2007 (doi:10.1038/nature06403) — local PDF verified end-to-end (NANOG bistability, germline requirement)
Loh 2006 (doi:10.1038/ng1760) — local PDF verified end-to-end (co-occupancy claim)
Chambers 2003 (doi:10.1016/s0092-8674(03)00392-1) — local PDF verified pages 1-6 (discovery, foundational claims)
Mitsui 2003 (doi:10.1016/s0092-8674(03)00393-3) — local PDF verified pages 1-6 (discovery, foundational claims)
Boyer 2005 (doi:10.1016/j.cell.2005.08.020) — download failed (status: failed); claim left unverified
Yu 2007 (doi:10.1126/science.1151526) — not_oa; closed access; unverifiable
Lapasset 2011 (doi:10.1101/gad.173922.111) — cross-checked against verified partial-reprogramming.md (no PDF re-read per SOP)
UniProt Q9H9S0 — REST API verified (length, homeodomain residues, WR repeats, transactivation regions)
Open Targets Platform ENSG00000111704 — API verified (all tractability flags false; tier 4 confirmed)
NCBI Gene 79923 — eutils confirmed
GenAge — NANOG absent (confirmed; appropriate)

Corrections made:

Co-occupancy claim: “>80% of OCT4-bound regions are also bound by NANOG” → “345 of 776 Oct4-bound genes (44.5%) also bound by NANOG within 50 kb of the transcription unit” (Loh 2006 Fig. 3b, text p.433). The original “>80%” figure does not appear in the paper; the Venn’s per-circle numbers are Oct4-only (431), overlap (345), Nanog-only (1,457) — giving 345/(431+345)=44.5% of Oct4-bound genes co-occupied, not 80%.
Method: “ChIP-chip” → “ChIP-PET (chromatin immunoprecipitation paired-end ditags)” for Loh 2006. The paper uses ChIP-PET; the wiki’s seeder conflated this with ChIP-chip (used in Boyer 2005 for human ESCs).
Footnote [^loh2006]: updated to include binding-site numbers (1,083 Oct4 sites, 3,006 Nanog sites) and the correct gene-level co-occupancy statistic (345/776 = 44.5%).
Druggability section: Removed incorrect note that API returned a 400 error; replaced with confirmed API result (all tractability flags false, tier 4 confirmed).
Removed ⚠️ auto-extraction banner.
Flipped verified: false → true.

Unverifiable claims:

Boyer 2005 human-ESC co-occupancy claim (download failed; status: failed — not not_oa, so retry possible)
Yu 2007 Thomson 5-factor cocktail factual content (closed access; widely replicated in secondary literature)

Downstream propagation needed:

None identified. No other wiki page contains the “>80% OCT4-bound regions” co-occupancy number sourced to Loh 2006 (the claim originated in this new page). Verify with grep before confirming.

Final state: verified: true (partial scope — Boyer 2005 and Yu 2007 unverified)

[2026-05-07] R31 — Epigenetic alterations deepening (DONE; 5/5 pages seeded + verified)

Round 31 status: COMPLETE. Triggered from the 2026-05-07 lint-pass plan. Pre-flight inventory revealed many epigenetic atoms already seeded (dnmt3a, sirt1/3/6, tet1/2/3, partial-reprogramming process, in-vivo-partial-reprogramming-therapy intervention, oct4/sox2/klf4/c-myc), so R31 focused on the heavyweight orphan study pages (Lu 2020, Yang 2023, Ocampo 2016, Horvath 2013) plus the HDAC class page.

Pages seeded + verified (5):

studies/lu-2020-osk-vision-restoration.md — Sinclair-lab AAV2-OSK retinal vision restoration (10 inbound)
studies/yang-2023-epigenetic-information-loss.md — Sinclair-lab ICE/I-PpoI epigenetic-loss-as-cause paper (9 inbound)
studies/ocampo-2016-partial-reprogramming.md — Belmonte-lab in-vivo cyclic OSKM in LAKI progeria + WT aged (7 inbound)
studies/horvath-2013-epigenetic-clock.md — foundational multi-tissue 353-CpG clock (7 inbound)
molecules/proteins/hdac.md — Zn²⁺-dependent HDAC family class page (HDAC1–11; 10 inbound)

R31 verification statistics:

5 pages verified end-to-end
~50 source-text-level corrections
~15 scientifically critical corrections, including:
- Yang 2023 verifier (CRITICAL — main-agent error): Yang 2023 uses 3-factor OSK throughout, NOT 6-factor OSKMNL as my seeder brief asserted. The brief’s confidently-stated OSKMNL claim was fabrication on the main agent’s part — propagated into the Yang 2023 page seed, then caught by the verifier reading the actual paper. The wiki’s existing pages (partial-reprogramming.md, in-vivo-partial-reprogramming-therapy.md, aav-osk.md, information-theory-of-aging.md) had previously been verified and correctly say “OSK” for Yang 2023 — the OSKMNL fabrication did not propagate beyond the new study page (which was corrected). The OSKMNL cocktail belongs to Lapasset 2011 (full-iPSC reprogramming of senescent cells), not Yang 2023.
- Yang 2023 verifier: I-PpoI cuts at 20 sites total (19 non-coding), not “17–20 non-coding”; clock acceleration uses 2 study-trained clocks, not 4; primary readout is DNAme clocks (transcriptomic only for RGC arm).
- Lu 2020 verifier: Aging arm ages “12 months” was WRONG → actually 3 months and 11 months (page 7 of paper); 12 mo appears only in transcriptomic Extended Data follow-up; vision restoration absent at 18 months due to corneal opacity. DNAm clock is NOT Horvath; the paper uses a custom 1,226-CpG RGC aging signature plus the Wang & Lemos 2019 ribosomal DNAm clock. Safety arm “21 months” is age-at-injection, not observation duration. AAV uses BOTH Tet-Off (primary ONC arm) and Tet-On (post-injury + aging arms), not Tet-On only. Human neuron arm and TDG (BER) mechanism were entirely absent from auto-extraction — added by verifier. Polycistronic delivery required (monocistronic ineffective).
- Ocampo 2016 verifier: ~33% lifespan extension was NOT stated as a percentage in the paper; it’s a curve-read estimate from Figure 4B (paper reports only log-rank Mantel-Cox p<0.0001). Teratoma claim was MAJOR ERROR: continuous OSKM in single-copy 4F mice caused rapid mortality within ~4 days from dedifferentiation-induced organ failure (NOT teratomas at 8 weeks); 8-week teratomas were observed only in two-copy 4F+rtTA mice on the cyclic protocol. Gene copy number — not dosing schedule alone — is the key safety determinant. WT experiment group sizes populated from figure legends (n=6 GTT, n=4 fiber CSA / Pax7+).
- Horvath 2013 verifier (CRITICAL — multiple wiki-wide claims fabricated): Cerebellum deceleration claim REMOVED — paper says “no significant difference between cerebellum and occipital cortex from same subjects.” HIV acceleration claim REMOVED — paper has no HIV mention; belongs to Gross 2016. Down syndrome acceleration claim REMOVED — paper has no DS mention; belongs to Horvath 2015 Genome Med. Cancer n: 6,000 → 5,826 (32 tissue datasets, 20 cancer types). Cancer age acceleration: 36 → 36.2 yr. Heritability: qualitative → 100% in newborns and 39% in older subjects (Falconer’s formula). Held-out test r=0.96 / MAE=3.6 yr distinguished from training (r=0.97 / 2.9 yr).
- HDAC verifier: Pao 2020 OGG1 K341 site attribution is wrong — the K341 site originates in Bhakat 2006 (Pao 2020 cites it as ref 47); Pao 2020 demonstrates HDAC1 deacetylates OGG1 on p300-acetylated sites, but doesn’t pin K341. Exifone behavioral data scope: contextual fear conditioning + LTP rescue was in 8-mo 5XFAD AD-model mice, NOT 17-mo aged WT (which only shows OGG1 activity rescue + 8-oxoG reduction). McDonald 2013 fly-lifespan drug identity: was sodium butyrate, actually vorinostat (SAHA) per McIntyre 2019 Table 1.
~5 wrong DOIs in seeder briefs caught and corrected (including 3 unverifiable ones rejected outright by HDAC seeder: my “Kane 2019 Nat Comm”, “Sun 2019 Genes Dev”, “Bhaskara 2008 Mol Cell” had no PubMed match)

Cross-cutting propagation done by main agent (R31 post-verification pass):

biomarkers/horvath-clock-2013.md — cancer-sample count corrected from “276 cancer (validation only)” → “5,826 cancer samples (32 tissue datasets, 20 cancer types; cancer-specific age-acceleration validation only — not training)”
processes/partial-reprogramming.md — Lu 2020 footnote rewritten with corrected ages (3 + 11 mo aging arm), TDG/BER mechanism added, custom RGC clock identity (not Horvath), polycistronic-required note, dual Tet-Off/Tet-On architecture; Teratoma-and-de-differentiation-risk section rewritten with the single-copy vs two-copy gene-copy-number distinction (Ocampo 2016 Fig 3E,F) — clarifying that continuous-dox single-copy 4F mice died from organ failure (~4 days), and 8-week teratomas are a two-copy phenomenon
hallmarks/epigenetic-alterations.md — partial-reprogramming section updated to wikilink the now-seeded study pages and the verified partial-reprogramming process page; clarification added that Lu 2020 + Yang 2023 use 3-factor OSK (not OSKMNL — OSKMNL is Lapasset 2011 context); the “Stub — not yet seeded” annotation on [[partial-reprogramming]] removed; study-references list at bottom of page updated to mark all 5 verified studies (Horvath/Waziry/Yang/Lu/Ocampo) with “verified R31” or “verified R27”

Schema escalations surfaced (for next CLAUDE.md cleanup):

Multi-arm n-subjects: for type: study — Lu 2020 has six experimental arms with no single headline N; seeder used n-subjects: null with explanatory inline comment. CLAUDE.md should formalize this case (e.g., allow free-text “varies by arm; see Design”).
Seeder-brief DOI memory unreliable persists: HDAC seeder rejected 3 of my asserted DOIs as not-real; consistent with R29 (16+ wrong DOIs caught) and R30 (10 wrong). Worth documenting more strongly in the user-facing memory feedback as an established pattern that has now appeared in 3 consecutive rounds.

Hallmark-level impact:

epigenetic-alterations mechanistic depth before R31: 12 prot / 2 path / 1 proc / 7 interv (per 2026-05-07 lint table; second-thinnest mechanism layer).
epigenetic-alterations mechanistic depth after R31: ~13 prot / 2 path / 1 proc / 7 interv (HDAC class page added; 4 heavyweight study pages now exist as standalone). The hallmark’s mechanism layer is still under-developed but the heavyweight evidence pages are now first-class citizens.

Remaining gaps in this cluster (deferred to R31b or later):

processes/dna-methylation.md + processes/dna-demethylation.md (6 inbound combined) — referenced from epigenetic-alterations.md key-pathways frontmatter but pages don’t exist
processes/histone-modification.md + processes/histone-acetylation.md (4-2 inbound) — referenced but absent
molecules/proteins/nanog.md (9 inbound) — pluripotency TF, distinct from oct4/sox2/klf4/c-myc which all exist
molecules/proteins/dnmt1.md (3 inbound) — maintenance methyltransferase
molecules/proteins/sirt2.md, sirt7.md — sirtuin family completeness (sirt1/3/6 exist)
interventions/pharmacological/hdac-inhibitors.md — class page (vorinostat, romidepsin, panobinostat, belinostat, chidamide; ricolinostat HDAC6-selective)
studies/schooling-2025-mr-epigenetic-clock-lifespan (5 inbound) — MR null result for clock causality
HDAC family member individual pages (HDAC1–11) — class page suffices for now; individual seeding only if specific aging-relevant primary literature accumulates

🧬 Aging Wiki

Explorer

R31

log/R31.md — Round 31 entries

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

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

[2026-05-07] verify — interventions/pharmacological/dna-pkcs-inhibitors.md

[2026-05-07] verify — molecules/proteins/mlh1.md

[2026-05-07] verify — interventions/pharmacological/parp-inhibitors.md

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

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

[2026-05-07] verify — molecules/proteins/msh2.md

[2026-05-07] verify — molecules/proteins/lig1.md

[2026-05-07] verify — molecules/proteins/brca2.md

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

[2026-05-07] verify — molecules/proteins/ape1.md

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

[2026-05-07] verify — molecules/proteins/xrcc1.md

[2026-05-07] R31b-β — Histone modification + sirtuin family + HDAC inhibitors (DONE; 5/5 pages seeded + verified)

[2026-05-07] verify — molecules/proteins/sirt7.md

[2026-05-07] verify — molecules/proteins/sirt2.md

[2026-05-07] verify — processes/histone-acetylation.md

[2026-05-07] verify — processes/histone-modification.md

[2026-05-07] verify — interventions/pharmacological/hdac-inhibitors.md

[2026-05-07] R31b-α — DNA methylation cluster (DONE; 5/5 pages seeded + verified)

[2026-05-07] verify — molecules/proteins/dnmt1.md

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

[2026-05-07] R31 — Epigenetic alterations deepening (DONE; 5/5 pages seeded + verified)

Graph View

Table of Contents

Backlinks