R42

log/R42.md — Round 42 entries

Sub-file of log — see parent for index.

[2026-05-19] R42 close — topical antioxidants + UV protection (6 pages)

Fifth round of the R38–R44 skin-aging coverage campaign. 6 atomic pages seeded + verified single-day. 13 agent invocations (6 seeders + 7 verifiers — niacinamide required retry after socket-closed error) + 1 propagation pass. All 6 pages flipped to verified: true (most with partial scope for closed-access primaries).

Pages added (all 6)

Compounds (5): molecules/compounds/{niacinamide, ascorbic-acid, alpha-tocopherol, dihydromyricetin, resveratrol}.md

Lifestyle intervention (1): interventions/lifestyle/uv-protection.md

Major verifier findings

R42 verifier yield was the highest of any R38–R44 round to date for full-PDF cross-checks. Multiple seeders pulled PDFs directly during verification (Falckenhayn 2024, Lin 2005, ATBC 1994, Krutmann 2021, Alvares 2022, Luze 2020, Shariff 2022, Ghasemiyeh 2025, plus all 4 resveratrol mechanism papers Howitz 2003 / Pacholec 2010 / Wood 2004 / Hubbard 2013). Verifier yield matched R40’s pattern of training-memory fabrication catches.

Canonical-DB-memory hallucinations (caught despite new feedback memory in place):

• Niacinamide ChEMBL CHEMBL1140 + DrugBank DB02701 — pulled correctly by seeder from PubChem REST (per feedback memory discipline ✓)
• Ascorbic acid ChEMBL CHEMBL196 vs PubChem-cross-ref CHEMBL40274 — resolved via direct ChEMBL API (CHEMBL40274 is structurally different isomer; CHEMBL196 is canonical)
• DHM CHEMBL CHEMBL3348861 — pulled directly from PubChem (per discipline ✓)
• Alpha-tocopherol natural + synthetic CIDs (14985 + 2116) — seeder caught the stereoisomer distinction and surfaced pubchem-cid-synthetic: schema escalation

DOI / PMID fabrications:

• Murad 1981 PMID WRONG (ascorbic acid): I asserted PMID 6268471; that’s an LH receptor paper. Correct PMID 6265920 + DOI 10.1073/pnas.78.5.2879. Verifier resolved.
• Chen 2015 NF-κB DOI WRONG (DHM): I asserted “Chen G, Tian X,… Mol Pharmacol”; not the right paper. Correct: Hou XL et al. J Nat Prod 2015, doi:10.1021/acs.jnatprod.5b00275, PMID 26171689.

Mechanism / quantitative fabrications:

• Ferulic acid metal-chelation fabrication (alpha-tocopherol): I asserted ferulic acid “chelates metal ions that catalyse ascorbate oxidation, stabilising the formulation”; Lin 2005 explicitly says “the mechanism of ferulic acid’s stabilizing effect on vitamins C and E is unknown” (p.830). Pure seeder-brief fabrication.
• Hubbard 2013 substrate FABRICATED (resveratrol): I asserted “1-Aminoanthracene specific peptide substrate”; actual paper uses PGC-1α Lys778 / FOXO3a Lys290 sequences. Verifier caught.
• Hubbard 2013 author Bhatt FABRICATED: “Bhatt” not in author list. Correct: Hubbard BP, Gomes AP, Dai H,…, Sinclair DA.
• DHM binding mode FABRICATED: I asserted “competitive or allosteric inhibition of catalytic domain”; Falckenhayn 2024 doesn’t characterize binding mode. Removed.
• DHM DNMT3A/3B target FABRICATED: I asserted broader target spectrum; paper has DNMT1-only assay. Removed.

Cohort / study-design fabrications:

• Shariff year + n + concentrations all WRONG (niacinamide): 2021 → 2022; n=42 split-face → n=44/50 enrolled (CONSORT); 5% niacinamide → 3% niacinamide; 2% 4-hexylresorcinol → 0.4% 4-hexylresorcinol. Plus Chinese-Shanghai-women-35-60 cohort.
• Hughes 2013 design incomplete (UV-protection): I described 2 arms; actual 2×2 factorial with β-carotene/placebo cross-arm; β-carotene arm omitted from wiki entirely.
• Hughes 2013 outcome measure WRONG (UV-protection): “Skin Surface Aging Scale” → “microtopography of dermal elastosis on the dorsal hands”; relative odds 0.76 (95% CI 0.59-0.98) added.
• Padayatty 2004 n + dose + fold-difference WRONG (ascorbic acid): n=7 → n=17; “18 g/day → 220 μmol/L” → “3 g every 4 hours → 220 μmol/L”; “25-fold higher” → “~60-fold higher” (13,400 vs 220 μmol/L).
• Humbert 2003 design WRONG (ascorbic acid): “split-face” → “split-body (low-neck and arms)”.
• Alvares 2022 unit confusion (UV-protection): n=40 forearms → n=40 participants (80 forearms).

Falckenhayn 2024 quantitatives added (DHM, missing from seeder): DNMT1 inhibition at 20 µM, n=13 donor lines, global hypomethylation p=1.4×10⁻³³ (101,067 differentially methylated probes); DNAm-age Δ −1.86 yr (Skin & Blood clock) + Δ −2.03 yr (epidermis clock). Vehicle-controlled in-vivo arm previously omitted entirely: n=30 female volunteers, topical 0.15% DHM, 2 weeks + UV challenge, 12,444 probes shifted toward young-epidermis pattern. This is the critical missing piece that resolves the previously-tagged #gap/no-mechanism for “does topical DHM reach basal keratinocytes” — Falckenhayn 2024 actually has a vehicle-controlled topical arm.

Supersession candidates surfaced

Niacinamide / ONTRAC story now nuanced:

• Allen 2023 NEJM (PMID 36856616; doi:10.1056/NEJMoa2203086): same 500 mg BID nicotinamide for 12 months in organ-transplant recipients (immunosuppressed); n=158; NO reduction in keratinocyte cancers (RR 1.0, 95% CI 0.8-1.3, p=0.96). ONTRAC’s 23% NMSC-reduction benefit does NOT generalize to immunosuppressed populations.
• Tosti 2023 systematic review + meta-analysis (4 trials): “insufficient evidence” pooled across immunocompetent + immunosuppressed populations.
• Niacinamide page now reflects this nuance with #gap/contradictory-evidence tag.

Resveratrol / STAC:

• Mansouri 2025 (PMID 40158656, JAND SR + GRADE meta-analysis, 11 RCTs) — concludes “resveratrol supplementation does not significantly influence human SIRT1.” Adds RCT-level evidence reinforcing the limited-negative framing per CLAUDE.md R24.

DHM:

• Bienkowska 2026 (Clin Epigenetics; n=851 population cohort; PMID 41957838) — DHM associated with decelerated epigenetic aging in skin (observational; same Beiersdorf group).

Propagation work completed

• frameworks/interventions-by-hallmark.md line 273: genomic-instability row “0 | low | No direct-mechanism intervention exists” → “1 (uv-protection) | low | UV-protection added R42 — first direct-mechanism intervention (UVB → AP-1/NF-κB → MMP / CPD photoaging axis); coverage gap partially resolved”. First R42 milestone: this hallmark previously had zero linked interventions wiki-wide.
• frameworks/intervention-classes.md line 899 niacinamide-in-tyrosinase-inhibition error fixed: niacinamide removed from tyrosinase-inhibition canonical list (correct mechanism is PAR-2 melanin-transfer inhibition, properly catalogued at the separate melanin-transfer-inhibition class entry).
• frameworks/intervention-classes.md pre-existing duplicate extracellular-matrix-remodeling class entries merged: line 942 (R36 2026-05-09 GHK-Cu-specific) deleted; copper-chaperone sub-mechanism note merged into the broader line 583 entry (2026-05-12; HA + GHK-Cu). Canonical values now: extracellular-matrix-remodeling, collagen-synthesis-stimulation, copper-chaperone.
• interventions/pharmacological/nad-precursors.md disambiguation added: reciprocal note pointing to topical [[niacinamide]] page; documents the topical-vs-systemic distinction + ONTRAC+Allen 2023 nuance.
• Round-number error fixed inline by ascorbic-acid verifier: 3 mechanism class entries labeled “R43” → “R42” in intervention-classes.md.
• molecules/compounds/resveratrol.md mechanism value standardized: sirtuin-activator-disputed → canonical sirtuin-activator (body retains the dispute framing).

Propagation confirmed not needed:

• Ferulic acid metal-chelation fabrication: only appears in log.md (correction record). ✓
• Padayatty 2004 fold-difference: only on ascorbic-acid page itself. ✓
• Shariff year/n/concentration errors: only on niacinamide page itself. ✓

New mechanism classes registered in intervention-classes.md (6 added during R42)

• DNMT-inhibitor (R42 from DHM seeder) — hallmarks: epigenetic-alterations
• lipid-peroxidation-inhibitor (R42 from alpha-tocopherol seeder) — hallmarks: mitochondrial-dysfunction, chronic-inflammation
• epidermal-barrier-repair (R42 from niacinamide seeder) — hallmarks: chronic-inflammation, loss-of-proteostasis
• melanin-transfer-inhibition (R42 from niacinamide seeder) — distinct from tyrosinase-inhibition
• UV-filter / photoprotection (R42 from UV-protection seeder) — hallmarks: genomic-instability, loss-of-proteostasis, chronic-inflammation
• prolyl-hydroxylase-cofactor + melanin-inhibition + tet-enzyme-cofactor (R42 from ascorbic-acid seeder, originally labeled R43 — corrected inline)

Schema escalations surfaced

• pubchem-cid-synthetic: field for type: compound (alpha-tocopherol precedent; parallel to R41 chembl-id-active-metabolite: from tazarotene). Both indicate a broader pattern: prodrug/stereoisomer compounds with biologically distinct active forms need parallel canonical-DB fields.
• mechanisms: field route-of-administration ambiguity (niacinamide precedent): niacinamide IS biochemically a NAD+ precursor but topical doesn’t engage that pathway. The mechanisms field doesn’t encode route. Document on individual compound pages.

Forward queue surfaced by R42

• studies/qi-2026-dhm-epigenetic-skin-aging.md PRIORITY VERIFICATION: study page is still verified: false despite being the initiator of the entire R38–R44 campaign. Should be prioritized for dedicated PDF verification. Falckenhayn 2024 PDF now downloaded — would aid the Qi 2026 verification.
• studies/falckenhayn-2024-dhm-dnmt-inhibitor.md — natural seeding target after PDF read during R42 DHM verification.
• studies/hughes-2013-sunscreen-photoaging.md — primary photoaging-prevention RCT; widely cited but no dedicated study page; closed-access verification but worth seeding.
• studies/lin-2005-ce-ferulic-photoprotection.md — Yorkshire pig model; PDF now local; CE+Ferulic mechanism anchor.
• molecules/compounds/ferulic-acid.md — referenced repeatedly across R42 pages (CE+Ferulic synergy); no dedicated page. Natural follow-on.
• Allen 2023 + Tosti 2023 ONTRAC supersession integration: niacinamide page now has the qualification; sirtuin-activators.md class page + sirt1.md may benefit from Mansouri 2025 RCT-level meta-analysis citation.
• molecules/proteins/par-2 (F2RL1): niacinamide’s canonical mechanism receptor; no protein page yet.

Round disposition

R42 closed cleanly. 6/6 pages flipped to verified: true. ROADMAP.md updated. Ready to dispatch R43 (skin epigenetic biomarkers + clinical-AGE imaging; 4 pages: Bormann-epidermis-clock-2016, Qi-23k-epidermis-clock-2026, TapeLift-clock-2026, skin-autofluorescence-AGE-reader).

---

[2026-05-19] verify — ascorbic acid compound page (R42)

Page verified: molecules/compounds/ascorbic-acid.md Sources: Canonical-DB IDs verified via PubChem REST + ChEMBL API + Crossref DOI checks; primary studies Humbert 2003 / Nusgens 2001 / Lin 2005 / Kawahori 2020 / Correia 2023 / Sanabria 2023 / Žmitek 2024 / Zi 2026 cross-checked via PubMed efetch or Crossref metadata; Paddayatty 2004 verified via PubMed abstract; Murad 1981 PNAS identified and corrected via PubMed search; Pinnell 2001 DOI confirmed via Crossref. Corrections applied: 7

1. Murad 1981 PMID: 6268471 (wrong — unrelated LH receptor paper) → 6265920 (correct PNAS prolyl hydroxylase paper); DOI 10.1073/pnas.78.5.2879 added; unsourced tag removed; footnote substantially updated.
2. Paddayatty 2004 n: “n=7” → “n=17 healthy hospitalized volunteers” (PubMed abstract).
3. Paddayatty 2004 oral dose ceiling: “maximum oral dose (18 g/day) achieves only 220 μmol/L” → “maximum tolerated oral dose 3 g every 4 hours → predicted peak 220 μmol/L” (actual paper value).
4. Paddayatty 2004 fold-difference: “25-fold higher” → “~60-fold higher (13,400 μmol/L [50g IV] vs 220 μmol/L [max oral])”; “57-64 μmol/L” claim removed (not from this paper).
5. Humbert 2003 study design: “split-face” → “split-body (low-neck and arms)” per abstract; n=20 changed to “n unconfirmed from abstract” (not stated in abstract).
6. Žmitek 2024 n: “n size in paper” → n=87 total (29 placebo, 28 CPHA, 30 CP; women 40–65 yr) confirmed from PubMed abstract; dose details added (collagen 5 g + vitamin C 80 mg ± HA 30 mg).
7. R43 label in intervention-classes.md: three mechanism class “Added:” labels for prolyl-hydroxylase-cofactor, melanin-inhibition, tet-enzyme-cofactor incorrectly read R43 → corrected to R42 (this seeder session was R42). Unverifiable claims:

• Humbert 2003 n=20 (cited in seeder brief; not confirmed from abstract — PDF pending)
• Nusgens 2001 n=13 (not stated in abstract — PDF pending)
• Pinnell 2001 formulation details (not_oa; verified against abstract only)
• Zi 2026 quantitative claims (NOT_IN_ARCHIVE)
• Sanabria 2023 (not_oa; no-fulltext-access)
• Kawahori 2020 suckling-pup n and specific 5hmC changes (abstract only; PDF pending) Supersession candidates: none found (PubMed search for topical vitamin C meta-analysis 2023–2026 returned no relevant new systematic reviews or large RCTs not already incorporated). Downstream propagation: none required — no other wiki pages cite ascorbic-acid study footnotes directly; intervention-classes.md R43→R42 correction applied inline.

[2026-05-19] verify — niacinamide compound page (R42 retry)

Page verified: molecules/compounds/niacinamide.md Sources: Shariff 2022 (doi:10.1111/ics.12759) PDF-verified; Ghasemiyeh 2025 (doi:10.1038/s41598-025-26693-8) PDF-verified; Chen 2015 ONTRAC (doi:10.1056/NEJMoa1506197) Crossref metadata-only (not_oa); canonical-DB IDs confirmed via PubChem REST + ChEMBL API; supersession search performed. Corrections applied: 6

1. Shariff n: “n=42 split-face” → “n=44 completing/50 enrolled” (PDF: 50 enrolled, 6 withdrew, 44 completed all phases)
2. Shariff niacinamide concentration: “5% niacinamide alone” → “3% niacinamide alone” (PDF: both arms used 3%; the active arm added 0.4% 4-HR to the 3% niacinamide base)
3. Shariff comparator: “2% 4-hexylresorcinol + 5% niacinamide vs 5% niacinamide alone” → “0.4% 4-hexylresorcinol + 3% niacinamide vs 3% niacinamide alone” (exact concentrations from PDF Methods)
4. Shariff population/year: Chinese women 35–60 years (Shanghai) added; publication year corrected from 2021 (epub) to 2022 (print: Int J Cosmet Sci 2022;44:103–117); author list corrected to full team (Damodaran A corresponding, Unilever R&D)
5. Ghasemiyeh n: “n=not stated in abstract” → “n=99 enrolled (33/group); n=77 completed 3-month treatment” (CONSORT flowchart in PDF)
6. Ghasemiyeh niacinamide concentration: concentrations now specified — niosomal arm TXA 2%/NCA 2%; conventional arm TXA 5%/NCA 4%; 3-month duration added Supersession candidates found (R25):

• Allen et al. 2023 NEJM (PMID 36856616; doi:10.1056/NEJMoa2203086) — oral nicotinamide 500 mg BID in transplant recipients; n=158; RR 1.0 (p=0.96) — no benefit; qualifies ONTRAC finding to immunocompetent populations only
• Tosti et al. 2023 Nutrients meta-analysis (PMID 38201930; doi:10.3390/nu16010100) — 4 trials pooled; RR NMSC 0.82 (ns) — insufficient evidence overall when immunosuppressed population included Supersession note added to ONTRAC section body with contradictory-evidence; flagged for main agent re-seeding decision Flags for main agent:
• intervention-classes.md line 899: niacinamide incorrectly listed as “canonical tyrosinase inhibitor” — niacinamide’s mechanism is PAR-2-mediated melanin transfer inhibition, not direct tyrosinase inhibition; correction needed on that page

[2026-05-19] verify — UV protection intervention page (R42)

Page verified: interventions/lifestyle/uv-protection.md Sources: Krutmann 2017 and Fisher 1996 (cross-checked against R39-verified study pages); Krutmann 2021 PDF (downloaded from Manchester Research OA during this pass); Alvares 2022 PDF (downloaded from PMC); Luze 2020 PDF (downloaded from PMC); Hughes 2013 (abstract via PubMed efetch + Krutmann 2021 secondary description; primary PDF closed-access). Corrections applied: 7

1. Hughes 2013 design: missing 2×2 factorial structure (4 arms) and beta-carotene null result — added
2. Hughes 2013 outcome measure: “Skin Surface Aging Scale” → “microtopography of dermal elastosis on dorsal hands”
3. Hughes 2013 p-value: “p-value not stated” → “relative odds 0.76 (95% CI 0.59–0.98)”
4. Alvares 2022 body text: “n=40 forearms” → “n=40 participants (80 forearms)”
5. Alvares 2022 footnote: corrected pages (72–79 → 157–165); full design details added; download status updated
6. Luze 2020 footnote: corrected pages (336–343 → 424–432); author Mülleger → Müllegger; methodology + conclusion quote expanded; download status updated
7. Krutmann 2021 section + footnote: updated to reflect actual paper content (SPF ≥30 + UVA-PF; 25% under-application; iron oxide evidence is melasma/actinic lentigo specific); download status updated Unverifiable: Rönsch 2021 (closed-access; no-fulltext-access added to footnote); Lyons 2021 (closed-access; no-fulltext-access added to footnote) Downstream update applied: krutmann-2017-skin-aging-exposome.md forward-ref stub removed (page now seeded+verified) Flags for main agent: interventions-by-hallmark.md genomic-instability count still shows 0 (propagation needed); intervention-classes.md duplicate extracellular-matrix-remodeling entries at lines ~583 and ~942 (pre-existing; cleanup pass needed)

[2026-05-19] verify — alpha-tocopherol compound page (R42)

Page verified: molecules/compounds/alpha-tocopherol.md

• Identity fields (PubChem CIDs 14985 + 2116, InChIKey GVJHHUAWPYXKBD-IEOSBIPESA-N, CAS 59-02-9, ChEMBL47, DB00163, molecular formula C29H50O2, MW 430.7 Da) confirmed via PubChem REST API — all correct
• Lin 2005 (doi:10.1111/j.0022-202x.2005.23768.x) PDF read end-to-end: concentrations confirmed (15% vit C + 1% vit E + 0.5% FA); porcine model confirmed; n=6 erythema/sunburn, n=3 Western blot added to footnote
• ATBC 1994 (doi:10.1056/NEJM199404143301501) PDF read end-to-end: n=29,133; median 6.1 years; α-tocopherol dose 50 mg/day; lung cancer −2% P=0.8; β-carotene +18% P=0.01 — all confirmed; footnote enriched with full quantitative detail
• HOPE-TOO (Lonn 2005, doi:10.1001/jama.293.11.1338) verified against PubMed abstract (PMID 15769967); n=9,541; RR for heart failure 1.13 (95% CI 1.01–1.26; P=0.03); all wiki numbers confirmed; full PDF not locally available (JAMA paywall)
• Corrections applied: 2
  1. Removed fabricated mechanism claim: wiki stated ferulic acid “chelates metal ions that catalyse ascorbate oxidation” — Lin 2005 explicitly states the stabilisation mechanism is unknown; corrected to reflect this + added no-mechanism tag
  2. Lin 2005 footnote: added experimental n values (n=6 / n=3), model species (Yorkshire pigs), removed “abstract-only” flag, noted mechanism-unknown finding
• Supersession check: no superseding RCT or meta-analysis for HOPE-TOO or ATBC findings post-2022; the null-to-harmful systemic vit E evidence base remains unchallenged
• literature-checked-through: 2026-05-19 (already set by seeder)
• Kayden 1993, Burton 1986, Bjelakovic 2013, Camillo 2022 are not OA; claims from these sources not full-PDF verified but are well-established and consistently hedged

[2026-05-19] verify — dihydromyricetin compound page (R42)

Page verified: molecules/compounds/dihydromyricetin.md

• Falckenhayn 2024 PDF read end-to-end (doi:10.3389/fragi.2023.1258184; downloaded from PMC during this pass; PMID 38500495)
• Canonical IDs confirmed via PubChem REST and ChEMBL API (all correct)
• Corrections applied: 7 (see verifier summary)
  1. DNMT1 mechanism section: replaced vague “micromolar concentrations” with specific dose (20 µM), n (13 donor keratinocyte lines), p-value (1.4×10⁻³³), and DNAm-age deltas (−1.86 yr / −2.03 yr)
  2. DNMT1 mechanism section: added the vehicle-controlled in-vivo arm (n=30) from Falckenhayn 2024 — was entirely omitted
  3. Removed unsupported “competitive or allosteric inhibition” binding mode claim
  4. Removed unsupported DNMT3A target from frontmatter and cross-references
  5. Chen 2015 NF-κB footnote: wrong authors (Chen G, Tian X, Mol Pharmacol) → corrected to Hou XL et al., J Nat Prod, doi:10.1021/acs.jnatprod.5b00275, PMID 26171689
  6. SIRT1/PRDX1 footnote: DOI resolved → doi:10.1016/j.phymed.2026.158160, authors Hu H et al.
  7. Falckenhayn footnote: PMID 38500495 added; archive status updated (downloaded); full study design n values added
• Supersession check: no superseding RCT/meta-analysis; Bienkowska 2026 (Clin Epigenetics, n=851 population-cohort) and Grönniger 2024 review are consistent with framing

[2026-05-19] verify — resveratrol compound page (R42)

Page verified: molecules/compounds/resveratrol.md

• PDFs read end-to-end: Howitz 2003 (nature01960), Pacholec 2010 (jbc.m109.088682), Wood 2004 (nature02789), Hubbard 2013 (science.1231097)
• PubChem CID 445154 / InChIKey confirmed via PubChem API
• Corrections applied: 5 (see verifier summary in main conversation)
• Hubbard 2013 footnote added (was cited in body prose, no footnote previously)
• mechanisms field reverted from non-canonical sirtuin-activator-disputed to sirtuin-activator
• verified: false → verified: true
• Supersession candidate flagged: Mansouri 2025 PMID 40158656 (meta-analysis, 11 RCTs, resveratrol does not significantly increase human SIRT1 expression)