R30

log/R30.md — Round 30 entries

Sub-file of log — see parent for index.

[2026-05-07] verify — schooling-2025-mr-epigenetic-clock-lifespan

Page verified: studies/schooling-2025-mr-epigenetic-clock-lifespan.md Source: Schooling CM et al. 2025 · doi:10.1186/s40246-025-00852-4 · Human Genomics 19:143 · local PDF read end-to-end (8 pp.)

Corrections (4):

1. GrimAge instrument count: “4–11 SNPs” → “4 SNPs” — wiki conflated GrimAge (4 SNPs) with PhenoAge (11 SNPs) in the exposures table; PDF Results + Fig. 2 caption confirm GrimAge n=4 for both lifespan and survival.
2. IEAA Men IVW lifespan CI upper bound: 0.10 → 0.06 — PDF Fig. 2 reads −0.04 (−0.15 to 0.06); wiki had 0.10 as the upper bound.
3. r² attribution error corrected — wiki stated epigenetic clocks explain ~3% of variance; PDF states 3% is for leukocyte telomere length, and epigenetic clock r² is described as “larger” (per-clock values in Supplementary Table 1 only). Corrected to attribute 3% to telomere; epigenetic clock r² noted as larger with Supplementary source.
4. MR-Egger intercept p-values for telomere length specified per-outcome — wiki cited (p=0.19 women; p=0.23 men) without outcome labels; these mix lifespan and survival outcomes (0.19 = women lifespan; 0.23 = men survival). All four intercept p-values now listed explicitly (men lifespan 0.16, women lifespan 0.19, men survival 0.23, women survival 0.55).

Additional qualification:

• I²_GX “>95.5% for all exposures” claim qualified — only the telomere I²_GX (96.5%) is stated in main text; per-clock values are in Supplementary Table 1 (not verified here). Sentence rewritten accordingly.
• GrimAge Limitations note expanded to clarify 4 instruments applies to both outcomes, with comparative context (PhenoAge 11, IEAA 23–24).

All other quantitative claims confirmed correct: sample sizes from Table 1 (472,174; 34,449/61/63/67; 415,311/412,937; 167,020/194,174), all IVW estimates and CIs from Fig. 2 (except IEAA Men upper CI above), all telomere estimates from Fig. 1, p=0.03 women survival direction/magnitude, MR-Egger significant intercepts (HannumAge F lifespan p=0.01; M survival p=0.03), author list (3 authors), journal/volume/article number, study-design metadata.

Downstream propagation needed (main agent):

• hypotheses/information-theory-of-aging.md — cites this study in key-evidence-against; verify that its framing of the null result is consistent with the corrections here (none of the corrections change the core null finding, so likely no change needed; worth a spot-check).
• processes/partial-reprogramming.md — cites this study as contradictory evidence for clock-reversal as therapeutic readout; same framing consistency check.

---

[2026-05-07] verify — hdac-class-page

Page verified: molecules/proteins/hdac.md Sources checked: 7 footnotes; 2 PDFs read in full (Pao 2020; Seto & Yoshida 2014); 1 full review read (McIntyre 2019); Choudhary 2009 verified via PubMed abstract (not_oa); Yang & Seto 2008 download failed; Glozak 2005 not_oa; McDonald 2013 download failed.

Corrections made (4):

1. K341 OGG1 site attribution: removed K341 from Pao 2020 claim — K341 originates from Bhakat 2006 (cited as ref 47 in Pao 2020); Pao 2020 identifies HDAC1 deacetylates OGG1 but does not report K341 itself. Added gap note.
2. Exifone behavioral claim scoped correctly: fear conditioning improvement and LTP shown in 5XFAD mice (not 17-month C57BL/6J wild-type); 17-month WT experiments showed OGG1 activity rescue and 8-oxoG reduction only.
3. McDonald 2013 drug identity corrected: wiki stated sodium butyrate; McDonald et al 2013 used vorinostat (SAHA). Sodium butyrate fly lifespan extension is Zhao 2005; phenylbutyrate is Kang et al 2002.
4. Footnote key corrected: [^mcdonald2012] → [^mcdonald2013] (DOI confirms 2013 publication year).

Unverifiable claims:

• Glozak 2005 (not_oa): non-histone substrate foundational claims unverified from full text.
• Yang & Seto 2008 (download failed): class taxonomy cross-confirmed against Seto & Yoshida 2014.
• McDonald 2013 (download failed): drug identity cross-confirmed against McIntyre 2019 Table 1.
• OGG1 K341 from Bhakat 2006 (not in wiki; not yet downloaded): flagged needs-replication.

Downstream propagation needed:

• molecules/proteins/ogg1.md — if it cites K341 from Pao 2020, that attribution needs correction (K341 is Bhakat 2006).
• Any page citing [[hdac]] for the behavioral rescue claim should note restriction to 5XFAD model.

[2026-05-07] verify — lu-2020-osk-vision-restoration

Page verified: studies/lu-2020-osk-vision-restoration.md Source: Lu Y et al. 2020, Nature 588(7836):124–129 · doi:10.1038/s41586-020-2975-4 · full PDF read end-to-end including main text + Extended Data Figs 1–10 + Methods (38 pages total).

Corrections made (8 categories):

1. Aging arm ages wrong: “~12 months old (primary endpoint) and ~11 months old” → 3 months old and 11 months old (paper text page 7). Vision restoration absent in 18-month-old mice.
2. DNAm clock identity wrong: “Horvath mouse clock” → custom 1,226-CpG RGC aging signature derived by the authors from their own RRBS data, plus ribosomal DNA clock (67/72 CpGs). Not the Horvath 2013 human clock.
3. Safety arm mischaracterized: “21 months of observation, 0 tumors” → (a) 10–18 months of continuous AAV9-OSK showed no excess tumor incidence; (b) Extended Data transgenic arm: 21-mo-old mice at injection start, 10 months dox, to ~32 mo age; background tumor rate 65–90% in all groups; no excess in OSK arm.
4. AAV architecture: wiki described only Tet-On throughout; primary ONC arm used Tet-Off (expression default-on, dox suppresses). Both systems present. Corrected.
5. Human neuron arm missing: entire Extended Data Fig. 8 section (15-fold neurite area at day 9 post-vincristine; TET2-dependent) was absent from auto-extraction. Added.
6. TDG mechanism missing: TDG knockdown abolishes OSK benefit (Extended Data Fig. 7). Added to mechanistic model.
7. ONC arm group sizes added from Extended Data Fig. 4a-b: n = 8 OSK / n = 6 control at 1 mo; polycistronic requirement documented.
8. COI refined: Sinclair is consultant to, patent licensor to, board member of, and equity owner of Iduna Therapeutics (Life Biosciences company); additional co-author disclosures added.

Downstream propagation needed (main agent):

• processes/partial-reprogramming.md — check [^lu2020] footnote for Horvath clock reference; check aging arm ages
• interventions/stem-cell-therapy/in-vivo-partial-reprogramming-therapy.md — check aging arm ages and safety arm duration
• hypotheses/information-theory-of-aging.md — check Lu 2020 DNAm clock attribution
• biomarkers/horvath-clock-2013.md — if Lu 2020 cited as using Horvath clock, incorrect

[2026-05-07] verify — yang-2023-epigenetic-information-loss

Pages verified: 1

• studies/yang-2023-epigenetic-information-loss — PDF read end-to-end (30 pages main text + methods). Major corrections:
  • CRITICAL: OSKMNL → OSK throughout. The paper exclusively uses 3-factor OSK (Oct4/Sox2/Klf4) for all reprogramming arms (in-vitro ICE cells, in-vivo whole-body AAV, RGC). OSKMNL does not appear in the paper. All 7 instances corrected.
  • I-PpoI cut sites corrected: “17–20 non-coding sites” → “20 total targets, 19 non-coding” (CTCTCTTAA▼GGTAGC at exactly 20 sites; none in mitochondrial DNA).
  • Clock arm clarified: ~50% acceleration uses 2 study-trained DNAme clocks (muscle + blood); 4 external clocks (Meer, Petkovich, Stubbs, Thompson) are the reversal arm. Primary readout for acceleration and reversal = DNAme; transcriptomics = RGC arm.
  • Aging phenotype quantified: FI at 10–12 mo post-treatment resembled 24-mo-old WT mice (p=0.0006 / p<0.0001); fear conditioning ~40% contextual recall; Barnes maze ~half of Cre; 6-fold fewer COX+ myofibers; kidney/skin/bone/cardiac histology added.
  • Induction protocol specified: 3-week tamoxifen; tissue Cre efficiency muscle 67%, liver 71%, hippocampus 61%, cortex 72%.
  • Mechanism corrected: SIRT1 + SIRT6 + HDAC1 + PARP1 (not SIRT1/HDAC1 alone); RCM label added.
  • c-Myc cross-reference corrected: OSK excludes c-Myc; c-Myc only in OSKM (Ocampo 2016).
  • Verified flag flipped; ⚠️ banner removed.

Downstream pages flagged for propagation (main agent):

• interventions/stem-cell-therapy/in-vivo-partial-reprogramming-therapy.md — likely contains OSKMNL for Yang 2023; correct to OSK
• processes/partial-reprogramming.md — Yang 2023 section likely describes OSKMNL; correct to OSK
• hypotheses/information-theory-of-aging.md — primary anchor; check OSKMNL in Yang 2023 claim blocks
• interventions/gene-therapy/aav-osk.md — may describe Yang 2023 as 6-factor; confirm OSK

[2026-05-07] verify — ocampo-2016-partial-reprogramming

Page verified: studies/ocampo-2016-partial-reprogramming.md Source: Ocampo et al. 2016, Cell 167(7):1719–1733.e12 · doi:10.1016/j.cell.2016.11.052 · full PDF read end-to-end including STAR Methods + all supplemental figures.

Corrections made (4):

1. Lifespan extension ~33% status clarified: paper does NOT state this percentage in the text. Reports “dramatic increase in median and maximal lifespan” with result shown only as Figure 4B survival curves. The ~33% is a visual estimate from the curves (medians ~18 vs ~24 weeks); now explicitly noted as a curve-read estimate, not a stated paper value.
2. Teratoma timing error corrected: wiki originally stated “continuous OSKM expression causes teratomas at 8 weeks.” Corrected to: (a) continuous dox in single-copy 4F mice caused rapid mortality within 4 days from organ failure — not teratomas; (b) teratomas at 8 weeks were from cyclic dox in two-copy (double-cassette) mice — not single-copy, not continuous. The safety characterization now correctly distinguishes gene-copy-number as the determining variable.
3. WT experiment group sizes added: replaced “~4–10 (verify)” with exact values from Figure 7 legends: GTT n=6/group; fiber CSA/central nuclei n=4/group; Pax7+ satellite cells n=4/group. Also added STZ dose (30 mg/kg, 5 daily injections) and CTX dose (10 µM, 50 µL IM).
4. Cyclic dosing description refined: introductory sentence in System 1 incorrectly implied continuous dox causes teratomas; corrected to accurately state continuous dox causes rapid mortality and teratomas belong to two-copy cyclic mice.

Verified unchanged (confirmed correct):

• Cyclic dosing schedule: 2 days ON / 5 days OFF — confirmed in STAR Methods
• Group sizes: LAKI −Dox n=20; LAKI +Dox n=13; LAKI 4F −Dox n=18; LAKI 4F +Dox n=15 — confirmed in Figure 4 legend
• n-subjects = 66 total (LAKI lifespan cohort) — confirmed
• Mouse strain: C57BL/6 background confirmed in Methods; G609G knockin confirmed in text
• WT aged mice: 12-month-old C57BL/6J — confirmed
• ECG group sizes n=4/group — confirmed Figure 4G
• Cyclic single-copy 4F mice: 35 cycles, no teratomas/dysplasia/pluripotency in any organ — confirmed
• Statistical test: log-rank (Mantel-Cox) — confirmed in Statistical Analysis section

Downstream pages to check (propagation for main agent):

• processes/partial-reprogramming.md (verified 2026-05-04): references Ocampo 2016 teratoma as avoiding teratoma via cyclic dosing — may need clarification that single-copy cyclic avoids teratoma, two-copy cyclic does not. The existing text says “cyclic dosing protocol reported no teratomas in the 2d-on / 5d-off protocol” — this is accurate for single-copy but should be qualified.
• interventions/stem-cell-therapy/in-vivo-partial-reprogramming-therapy.md (verified 2026-05-06): body mentions cyclic OSKM avoiding teratoma — may inherit the same qualification need.

[2026-05-07] R30 — Mitochondrial dynamics + biogenesis (DONE; 5/5 pages seeded + verified)

Round 30 status: COMPLETE. Triggered from the 2026-05-07 lint-pass plan. R30 was originally scoped as ~11 pages but pre-flight inventory revealed most “missing” pages (drp1, opa1, mitofusins, tfam, pgc-1alpha, mitochondrial-biogenesis, mitochondrial-dynamics) were already seeded + verified — the high inbound counts for mfn1/mfn2/mitochondrial-dynamics-pathway were alias-resolution false-positives in the inbound-count discovery script (mitofusins.md covers MFN1+MFN2 via aliases; mitochondrial-dynamics.md is the actual pathway page). Real R30 scope = 5 truly-missing pages.

Pages seeded + verified (5):

• processes/mitohormesis.md — adaptive response to mild mitochondrial stress; covers AMPK→PGC-1α→Nrf2 axis, mtUPR/ATFS-1/ATF5/DELE1-HRI arm, mitokines (MOTS-c/humanin/GDF15/FGF21), antioxidant paradox (Ristow 2009)
• processes/mtdna.md — mtDNA biology: 16,569 bp, 37 genes, replication via POLG/Twinkle, copy-number decline with age, mtDNA→cGAS-STING→IRF3 inflammaging axis (West 2015), MOTS-c (Lee 2015). Schema gap escalated: CLAUDE.md has no type: genomic-locus; type: process chosen as best fit. Defers heteroplasmy mechanism to existing [[mtdna-heteroplasmy]] page.
• molecules/proteins/nrf1.md — Nuclear Respiratory Factor 1; mitochondrial-biogenesis master TF; downstream targets TFAM/POLG/Twinkle/ETC subunits/TOMM/TIMM; PGC-1α coactivation via direct domain interaction (PGC-1 aa 180–403 with NRF-1 aa 108–143)
• molecules/proteins/nrf2.md — antioxidant master regulator (NFE2L2); KEAP1-CUL3 ubiquitination + Cys151 sensor; ARE target battery; druggability tier 1 with omaveloxolone (FDA-approved Feb 2023 for FA), bardoxolone BEACON failure, sulforaphane/DMF; cancer paradox (~17–20% NSCLC KEAP1 LOF)
• molecules/proteins/miro2.md — RHOT2 mitochondrial trafficking GTPase, sister to Miro1 (~60% identity); broader tissue distribution; Myo19/actin-based perinuclear distribution role; PINK1-Parkin axis substrate

R30 verification statistics:

• 5 pages verified end-to-end
• ~40 source-text-level corrections
• ~10 scientifically critical corrections, including:
  • mtDNA verifier: MOTS-c pathway DIRECTION REVERSED (was “AMPK → AICAR → ZMP”; actually folate-methionine cycle inhibition → AICAR accumulation → AMPK activation); West 2015 downstream effector corrected (NF-κB → IRF3); Larsson 1998 KO phenotype attribution corrected (germline lethal at E8.5–E10.5; cardiac DCM phenotype belongs to Wang 1999, not Larsson 1998)
  • mitohormesis verifier: Houtkooper 2013 organism list corrected (was “C. elegans, Drosophila, mouse”; actual paper covers worm + mouse hepatocytes only — no Drosophila); UPR^mt longevity is independent of ROS (Houtkooper 2013); DELE1-HRI mechanism mis-attributed to Durieux 2011 (correct: Fessler 2020 / Guo 2020 Science 370); Ristow 2009 p-value precision (ANOVA p<0.001 not p<0.05); Tapia 2006 footnote located and added (PMID 16242247)
  • NRF1 verifier: PGC-1α/NRF1 mechanism wrong (was “recruits CBP/p300 + SRC-1”; actually direct physical interaction PGC-1 aa 180–403 with NRF-1 aa 108–143 — CBP/p300 recruitment is Puigserver 1998 PPARγ context, NOT NRF-1); POLG/Twinkle attribution corrected; Huo 2001 mtDNA depletion quantified (30 → <5%); Wu 1999 quantitative results added (mtTFA promoter ~4-fold; mtDNA +80%)
  • NRF2 verifier: MOXIe trial label CORRECTED (“Phase 3” → “registrational Phase 2”); Pergola 2011 endpoint timepoint INVERTED (was 52 weeks, actually 24 weeks primary); BEACON HR added (1.83 for HF hospitalization); Lynch 2019 design corrected (Phase 2 randomized 3:1, not Phase 1/2); Neh7 attribution Yamamoto 2018 → Wang 2013; KEAP1 NSCLC mutation rate refined (~17–20%)
  • MIRO2 verifier: Birsa 2014 K27 ubiquitination MIS-ATTRIBUTED to MIRO2 (paper studies Miro1 throughout; MIRO2 mentioned once in passing); MIRO2-null mouse phenotype WRONG (was “lethal/severe”; actually viable and fertile to adulthood); Fransson 2003 NTPase activity claim unsourced (paper is sequence-analysis only); sequence identity 59→60%
• ~10 wrong DOIs in seeder briefs caught and corrected (MIRO2 alone: 5, including Fransson 2003, López-Domènech 2018, Birsa 2014, Klosowiak 2013, Modi 2019; plus Evans & Scarpulla 1989→1990 year fix on NRF1; consistent with seeder-brief-DOI-memory-unreliable pattern from R29)

Cross-cutting propagation done by main agent (R30 post-verification pass):

• hallmarks/mitochondrial-dysfunction.md — [[mtdna]] implicit-stub annotation closed (verified R30) with cGAS-STING→IRF3 axis added; [[oxphos]], [[mitochondrial-biogenesis]] annotations closed; [[drp1]] / [[mitofusins]] / [[opa1]] “planned stub” line replaced with verified-state list; [[mitohormesis]] cross-link upgraded to verified R30 with mtUPR-vs-ROS independence note; cross-references list expanded to include all 5 R30 pages plus disambiguating note that NRF1 ≠ Nrf2/NFE2L2.
• molecules/proteins/miro1.md — MIRO2 paralog description updated: 59% → 60% identity; clarified that K27-chain Parkin-Ser65 mechanism (Birsa 2014) was demonstrated for Miro1 only — MIRO2 chain-type unverified; Miro2-null mouse viability + Miro DKO E10.5 lethality added; Myo19/actin-based distribution role for Miro2 noted.
• pathways/pink1-parkin-pathway.md — Parkin substrate table: MIRO row updated to wikilink both [[miro1]] and [[miro2]]; mechanism column updated with neuron vs non-neuronal distribution distinction; Birsa 2014 K27/Ser65 mechanism scope-restricted to Miro1.
• molecules/proteins/drp1.md — [[mitochondrial-biogenesis]] “implicit stub” annotation closed (already verified — annotation drift).

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

• Type: genomic-locus — mtDNA forced into type: process because no type: genome or type: genomic-locus exists in CLAUDE.md. Applies also to: telomere DNA structure (R29 carryover), centromeric DNA, ribosomal-DNA arrays. Worth a dedicated type if more genomic-locus entities accumulate.
• NRF1 vs NRF2 disambiguation — three distinct genes share NRF/NFE2L abbreviations: NRF1 (mitochondrial biogenesis TF), NFE2L1 (proteasome regulator), NFE2L2/Nrf2 (antioxidant master). Pages cross-disambiguate in TL;DR; CLAUDE.md should formalize a “disambiguation block” convention for proteins with confusable names.

Hallmark-level impact:

• mitochondrial-dysfunction mechanistic depth before R30: 28 prot / 9 path / 7 proc / 22 interv (already well-covered per lint table).
• mitochondrial-dysfunction mechanistic depth after R30: ~31 prot / 9 path / 9 proc / 22 interv. R30 added biogenesis-master + antioxidant-master + trafficking-paralog + mtDNA-biology + mitohormetic-process — the mechanism layer is now near-complete for this hallmark.

Remaining gaps in this cluster (deferred):

• Cardiolipin (3 inbound) — endogenous IMM-specific lipid; schema decision needed (compound vs metabolite directory)
• HRI / EIF2AK1 (5 inbound) — ISR kinase, mostly proteostasis-relevant
• MFF, FIS1 — DRP1 receptors; lower priority since DRP1 page covers them by reference
• POLG2 (2 inbound) — POLG accessory subunit
• TOMM20, TIMM17A (2 each) — translocator subunits

---