🧬 Aging Wiki

R41

18 min read

log/R41.md — Round 41 entries

Sub-file of log — see parent for index.

[2026-05-19] R41 close — retinoid cluster + class MOC (7 pages)

Fourth round of the R38–R44 skin-aging coverage campaign. 7 atomic + MOC pages seeded + verified single-day. 15 agent invocations (7 seeders + 8 verifiers — tretinoin required retry after socket-closed error) + 1 propagation pass. All 7 pages flipped to verified: true (5 with partial scope for closed-access primaries).

Pages added (all 7)

Compounds (6): molecules/compounds/{tretinoin, retinol, retinaldehyde, adapalene, bakuchiol, tazarotene}.md

Class MOC (1): interventions/pharmacological/retinoids.md

Major verifier findings

R41 verifier yield was unprecedented for canonical-database errors specifically. 6 separate canonical-DB-memory hallucinations caught across 6 different pages, in addition to the usual DOI + citation corrections. This is the third consecutive round (after R39 + R40) showing the seeder-DOI-memory pattern extending to canonical-database identifiers (ChEMBL, CAS, UniProt accessions).

Canonical-DB-memory hallucinations (newly documented as [[feedback_canonical_db_memory_unreliable]]):

  • Adapalene CAS WRONG: my brief asserted 105392-23-2; actual 106685-40-9.
  • Bakuchiol ChEMBL WRONG: CHEMBL2364604 resolves to buparlisib hydrochloride (a PI3K inhibitor); correct CHEMBL262344.
  • Tazarotene ChEMBL parent WRONG: CHEMBL1213 resolves to 5,6-dihydroxy-8-aminoquinoline; correct CHEMBL1657.
  • Tazarotene active-metabolite ChEMBL WRONG: CHEMBL83991 resolves to an unrelated C20H24N6 compound; correct CHEMBL1201375.
  • Tazarotenic acid MW WRONG: 303.40 → 323.41 g/mol (correct C19H17NO2S formula).
  • Retinaldehyde ChEMBL WRONG: CHEMBL897 resolves to probenecid (a uricosuric gout drug); correct CHEMBL81379.

Pattern: wrong canonical IDs frequently resolve to real but unrelated compounds rather than non-existent IDs — making the error harder to detect without direct API lookup.

DOI hallucinations:

  • Fisher 1996 DOI WRONG (retinoids MOC): 10.1038/nm1196-694 → 10.1038/379335a0 — nm prefix is Nature Medicine, NOT Nature. The paper is Nature 379:335-339.
  • Draelos 2023 DOI WRONG (tazarotene): 10.1111/jdv.18701 resolves to a totally unrelated 2022 JEADV paper on cutaneous melanoma metastases; correct 10.1080/09546634.2023.2166346. First author + journal also wrong in same brief (“Stein Gold L” + JEADV → Draelos ZD + J Dermatol Treat).

Cohort / n-value fabrications:

  • Bagatin 2018 n WRONG (adapalene): I asserted n=80; actual n=128 ITT (65 + 63) with PP=114.
  • Fisher 1996 n WRONG (retinoids MOC): I asserted “n=26”; correct “n=6-17 varies per experiment”.
  • Diridollou 1999 n + duration + p-value WRONG (retinaldehyde): n=44 → 40; 6mo → 1 year; p<0.05 → p<0.01.
  • Kim 2021 n + duration + design WRONG (retinaldehyde): not specified → n=23; 12 weeks → 8 weeks; “vs vehicle” → split-face vs retinol.
  • Deda 2026 cohort missing (retinaldehyde): added n=56 women, ages 30-58, 24 weeks (from Crossref).

Geographic / site fabrications:

  • Fisher 1996 skin site WRONG (retinoids MOC + retinol): “forearm” → buttock skin (the canonical R39-verified study page says buttock; “forearm” appears to be a persistent training-memory error for Fisher-lab studies — caught earlier in R38 Langerhans cells verifier).

Subtype-attribution fabrications:

  • RAR-Îł keratinocyte / RAR-β fibroblast claim (retinoids MOC): I asserted this attributed to Fisher 1996 in the brief; verifier confirmed Fisher 1996 contains NO RAR subtype characterization. De-attributed + tagged #gap/needs-replication.
  • Retinol “5-20x less potent than tretinoin” attributed to Mambwe 2024: verifier read Mambwe 2024 PDF in full and confirmed the figure doesn’t appear there. Reframed as cosmeceutical-literature folk knowledge with #gap/dose-response-unclear.
  • ALDH1A2 skin-predominance attribution to Mambwe 2024 (retinol): verifier read PDF; claim absent. Flagged #gap/needs-replication.

Framing / direction errors:

  • Lin 2025 coarse-wrinkle ranking WRONG (tazarotene): I asserted tazarotene “ranked highest for coarse wrinkles”; actual paper says **retinol ranks #1, tretinoin #2, tazarotene 3 (OR 1.84). Major framing inversion.
  • Lin 2025 retinoid-class framing tightened (adapalene): full PDF read; isotretinoin dominates (OR 116.23); adapalene OR 9.29 did not reach significance (p=0.0554, only n=57 in NMA nodes).
  • Lin 2025 hyperpigmentation framing (tretinoin): I asserted “tretinoin superior for hyperpigmentation”; actual paper says “tretinoin AND retinol superior”.
  • Adapalene tolerability claim overstated: I said “markedly less severe retinoid dermatitis”; Bagatin 2018 found no statistically significant AE difference (302 vs 320, p=0.495) at these concentrations.

Author-list errors:

  • Fanning 2024 (bakuchiol): missing McGee SA (second author); corrected to Fanning JE, McGee SA, Ibrahim OI.
  • Lau 2024 (bakuchiol): missing two authors; corrected to Lau M, Mineroff Gollogly J, Wang JY, Jagdeo J.
  • Sumita 2018 journal WRONG (retinoids MOC): I said “Journal of Dermatological Treatment”; actual J Eur Acad Dermatol Venereol.

Mechanism / cohort discordances integrated:

  • Chien 2022 JAMA Dermatology MMP-2 finding (tretinoin) flagged as #gap/contradictory-evidence: paper claims MMP-2 mRNA reduction by retinoid precursors correlating with fine-wrinkle improvement, contradicts Fisher 1996’s explicit “MMP-2 NOT induced/modulated” claim. Different experimental context (precursors, not direct UV-tretinoin exposure); both views represented.
  • Fanning 2024 systematic review (bakuchiol): 12/15 bakuchiol trials lacked methodological controls (no blinding/randomization). Reframes the field from single-study Dhaliwal 2019 optimism to “high risk of bias” — page now properly cautions against single-study framing.
  • Siddiqui 2024 (retinoids MOC): 13/25 comparators equivalent to tretinoin (not universally inferior) — softens the “tretinoin uniquely best” framing.
  • Kumari 2025 Arch Dermatol Res NMA (PMID 39921709; tretinoin recency surfaced): found “0.1% tretinoin cream shows the lowest effect size” for hyperpigmentation among 7 RCTs (n=337) — potentially contradicts Lin 2025’s tretinoin-superior framing. Tagged #gap/contradictory-evidence. Likely explained by formulation heterogeneity (7 vs 23 RCTs) but worth follow-up.

Propagation work completed

  • Duplicate class registration in frameworks/intervention-classes.md reconciled: tretinoin seeder added ### rar-agonist (line 1334) lowercase; retinoids MOC seeder added ### topical retinoid / RAR agonist (line 408) title-case. Deleted the line-1334 duplicate; standardized all 7 R41 pages to use title-case canonical values (RAR-agonist, AP-1-transrepression, MMP-1-suppression, collagen-synthesis-stimulation) per the line-408 canonical class entry.
  • molecules/compounds/retinol.md Fisher 1996 footnote: “forearm skin” → “buttock skin (sun-protected; NOT forearm)” with explicit qualifier + AP-1 transrepression mechanism caveat.
  • Mechanism field casing standardized across tretinoin, retinol, retinaldehyde (were lowercase) — now title-case matching adapalene + tazarotene + retinoids MOC.

Propagation work confirmed not needed:

  • Fisher 1996 10.1038/nm1196-694 hallucinated DOI: grep wiki-wide shows it only in log.md entries (correction records, not active citations). All compound pages used the correct 10.1038/379335a0 (seeder briefs differed from the retinoids MOC seeder brief). âś“
  • Fisher 1996 “forearm” attribution: only on retinol.md (now fixed). Other compound pages didn’t carry it.

New feedback memory

added — documents the canonical-DB-memory hallucination pattern (ChEMBL, CAS, UniProt accessions wrong from training memory; resolve to unrelated real compounds, not 404s) and the recommended discipline (don’t assert canonical IDs in seeder briefs from memory; provide PubChem CID + name + have seeder pull cross-reference IDs). Builds on existing feedback_seeder_brief_doi_unreliable memory.

Schema escalations surfaced

  • chembl-id-active-metabolite: field for type: compound (tazarotene precedent) — prodrug compounds with distinct active-metabolite ChEMBL IDs are a common pattern (statins, NSAIDs, ester retinoids). Worth queuing for next CLAUDE.md cleanup.

Forward queue surfaced by R41

  • Kumari 2025 Arch Dermatol Res hyperpigmentation NMA (PMID 39921709) — surfaces a potential supersession candidate for any wiki page citing tretinoin’s hyperpigmentation efficacy specifically (Lin 2025 says tretinoin+retinol superior; Kumari 2025 says 0.1% tretinoin has lowest effect size). Heterogeneity explanation likely (7 vs 23 RCTs) but worth follow-up.
  • Fanning 2024 systematic review (bakuchiol; PMID 38564402) — established that 12/15 bakuchiol trials lacked methodological controls; should propagate this framing to any future bakuchiol-class compound pages.
  • molecules/proteins/rara, rarb, rarg protein pages — all R41 compound pages reference these but no pages exist. Natural seeding targets.
  • molecules/proteins/aldh1a1, aldh1a2, aldh1a3 — RALDH enzyme family for retinoid conversion. Surfaced by retinaldehyde page.

Round disposition

R41 closed cleanly. 7/7 pages flipped to verified: true. ROADMAP.md updated. Ready to dispatch R42 (topical antioxidants + UV protection; 7 pages: niacinamide, ascorbic-acid, alpha-tocopherol, dihydromyricetin, resveratrol, uv-protection lifestyle intervention).

[2026-05-19] verify — molecules/compounds/retinaldehyde.md (R41 retinoid cluster)

Pages verified: 1

  • molecules/compounds/retinaldehyde.md — verified: true (partial scope; all papers verified against PubMed abstracts; closed-access papers Creidi 1999, Pechère 2002, Kaya 2005, Sorg 2014, Siddiqui/Soleymani 2024 limited to abstract-only; open-access PDFs Diridollou 1999, Rouvrais 2017, Kim 2021, Kaya 2023, Deda 2026, Mukherjee 2006, Barnes 2010 triggered for download but remained pending; identity fields verified against PubChem CID 638015 and ChEMBL API directly)

Corrections applied (9):

  1. ChEMBL ID WRONG → FIXED: CHEMBL897 → CHEMBL81379 (CHEMBL897 is Probenecid, a gout drug; confirmed via ChEMBL API; correct ID CHEMBL81379 confirmed as “RETINAL” with InChIKey NCYCYZXNIZJOKI-OVSJKPMPSA-N)
  2. Diridollou 1999 n WRONG → FIXED: n=44 → n=40 (21 retinaldehyde, 19 control; per PubMed PMID 10473959 abstract)
  3. Diridollou 1999 duration WRONG → FIXED: 6 months → 1 year (per PMID 10473959 abstract: “1-year study”)
  4. Diridollou 1999 design overclaimed → CORRECTED: “double-blind, placebo-controlled” → “controlled prospective study (emollient control; not explicitly double-blind per abstract)”; p-value corrected from p<0.05 to p<0.01
  5. Kim 2021 n MISSING → FILLED: N/A → n=23 (per PMID 33569865 abstract)
  6. Kim 2021 duration WRONG → FIXED: 12 weeks → 8 weeks (per PMID 33569865 abstract)
  7. Kim 2021 comparator WRONG → FIXED: described as “vs vehicle control” → actually “vs retinol (same concentrations), active comparator split-face design”
  8. Deda 2026 n MISSING → FILLED: N/A → n=56 women (ages 30–58) (per Crossref metadata doi:10.1111/srt.70326)
  9. Deda 2026 duration WRONG → FIXED: 12 weeks → 24 weeks (per Crossref metadata)
  10. Soleymani 2024 first-author WRONG → FIXED: “Soleymani T et al.” → “Siddiqui Z et al.” (actual first author confirmed via Crossref doi:10.1007/s40257-024-00893-w; footnote label [^soleymani2024] retained for link integrity)
  11. Pechère 2002 description underclaimed → EXPANDED: described as “in vitro” only; abstract confirms in vitro + 2 clinical trials (one n=22); evidence level updated to “Limited (in vitro + small clinical trial n=22)”
  12. Rouvrais 2017 description underdescribed → UPDATED: n and duration added (n=36, 2 months); confirmed open-label (no placebo control arm); pages corrected from 119–128 to 35–42

Supersession check (R25): one 2025 RCT found (PMID 41312593; n=66; retinaldehyde cream as adjunct to aesthetic procedures) — does not supersede foundational monotherapy evidence. No meta-analysis found. literature-checked-through: 2026-05-19 confirmed.

Unverifiable claims:

  • The specific “3–5Ă— more potent than retinol, 3–5Ă— less than tretinoin” quantitative ratios attributed to Mukherjee 2006 are not confirmed in the abstract; Mukherjee 2006 establishes qualitative hierarchy but abstract does not state specific ratios. Claim retained with qualifier pending full PDF verification (download in progress).
  • Barnes 2010 (CD44 hyaluronate synergy; PLoS ONE) — PDF download pending; not_oa check not needed (gold OA), awaiting download completion.

Downstream propagation needed:

  • None required. No downstream pages inherit the corrected Diridollou n, Kim 2021 design details, Deda 2026 n/duration, or Pechère evidence tier from retinaldehyde.md specifically. The wrong CHEMBL897 was only on retinaldehyde.md (not propagated to other compound pages in the R41 cluster). Siddiqui/Soleymani attribution error was limited to retinaldehyde.md; retinoids.md already correctly uses Siddiqui Z et al. independently.

[2026-05-19] verify — molecules/compounds/tazarotene.md (R41 retinoid cluster)

Pages verified: 1

  • molecules/compounds/tazarotene.md — verified: true (partial scope; Fisher 1996 + Lin 2025 read in full from local PDFs; Draelos 2023 confirmed via PubMed abstract; Kang 2001/2005, Phillips 2002, Chandraratna 1997 are not_oa; claims from those sources verified against published abstracts + FDA prescribing information only)

Corrections applied (7):

  1. ChEMBL parent ID WRONG → FIXED: CHEMBL1213 → CHEMBL1657 (CHEMBL1213 is 5,6-dihydroxy-8-aminoquinoline, an unrelated compound; confirmed via ChEMBL API name search)
  2. ChEMBL active metabolite ID WRONG → FIXED: CHEMBL83991 → CHEMBL1201375 (CHEMBL83991 is an unrelated compound; tazarotenic acid = CHEMBL1201375, C19H17NO2S per ChEMBL API)
  3. CAS number null → FILLED: null → 118292-40-3 (confirmed via PubChem CID 5381 xrefs)
  4. Tazarotenic acid MW WRONG → FIXED: 303.40 g/mol → 323.41 g/mol (C19H17NO2S calculated; ChEMBL1201375 confirms 323.42)
  5. Draelos 2023 DOI WRONG → FIXED: doi:10.1111/jdv.18701 → doi:10.1080/09546634.2023.2166346 (the old DOI resolves to a 2022 melanoma metastases paper; confirmed via Crossref)
  6. Draelos 2023 first author WRONG → FIXED: “Stein Gold L” → “Draelos ZD”; journal corrected to J Dermatol Treat 2023;34(1):2166346 (not J Eur Acad Dermatol Venereol); confirmed via PubMed PMID 36622889
  7. Lin 2025 coarse-wrinkle framing OVERSTATED → CORRECTED: “ranked highest for coarse wrinkle improvement” is contradicted by the actual rankogram (retinol #1 by cumulative probability; tazarotene #3); corrected with full OR + CI data from Table 1 and Fig 3C. Fisher 1996 n values clarified: n=7–10 for retinoid pretreatment experiments (Fig 4); n=6–17/timepoint cited in old footnote was from the UVB induction experiments (Figs 1–2).

Supersession check (R25): only two post-2022 papers on tazarotene + photoaging found (Lin 2025 NMA already integrated; Sadick 2023 narrative review, not superseding). No new large RCT or meta-analysis supersedes existing evidence. literature-checked-through: 2026-05-19 confirmed.

Downstream propagation needed:

  • Any page citing “Stein Gold 2023” or doi:10.1111/jdv.18701 for the tazarotene/adapalene irritation trial should be updated to Draelos ZD 2023, doi:10.1080/09546634.2023.2166346.
  • interventions/pharmacological/retinoids.md — verify “ranks highest for coarse wrinkle” framing is not present; if so, apply the same Lin 2025 correction.

Schema note: chembl-id-active-metabolite: is a non-standard frontmatter field. Other ester prodrugs may warrant the same field. Queue for CLAUDE.md escalation review.

[2026-05-19] verify — molecules/compounds/adapalene.md (R41 retinoid cluster)

Pages verified: 1

  • molecules/compounds/adapalene.md — verified: true (partial scope; Fisher 1996, Bagatin 2018, Kim/Rosenfield 2000 PDFs read end-to-end; Lin 2025 NMA PDF pages 1–6; Tran 2025 from Crossref full abstract only — PDF download failed)

Corrections applied (6):

  1. CAS number wrong → CORRECTED: 105392-23-2 → 106685-40-9 (confirmed via PubChem CID 60164 synonyms; both in identity table and frontmatter)
  2. Bagatin 2018 n wrong → CORRECTED: n=80 → n=128 ITT (65 adapalene + 63 tretinoin); PP=114 (57 per arm); table, footnote, comparison table, and limitations section all updated
  3. Tolerability claim overstated → CORRECTED: “markedly less severe with adapalene than tretinoin” → Bagatin 2018 found similar AE burden (302 vs 320 AEs, p=0.495); tolerability advantage is theoretical/class-level, not confirmed in this RCT at 0.3% adapalene vs 0.05% tretinoin
  4. Lin 2025 NMA retinoid-class-top-ranked claim → CORRECTED: isotretinoin ranked highest (OR 116.23); adapalene OR 9.29 (p=0.0554, NS in random-effects model); adapalene not statistically significant at NMA level; added actual ORs with CIs; removed needs-full-text-verification tag
  5. Bagatin 2018 footnote n=80 → CORRECTED to n=128 ITT with full efficacy and AE statistics
  6. Lin 2025 footnote abstract-only → UPDATED with actual NMA results from PDF

Supersession check (R25): Lin 2025 NMA (23 RCTs, 3905 participants) is the most current synthesis — already incorporated. Tran 2025 RCT (n=58) is most recent standalone RCT — already incorporated. Sadick 2023 review (PMID 37683070) is a narrative review, does not supersede efficacy claims. No superseding large RCT or meta-analysis that changes the core framing. literature-checked-through: 2026-05-19 confirmed.

Downstream pages that may need updates:

  • None identified — adapalene.md has no verified downstream entity pages citing it yet; the retinoid cluster MOC page (if seeded) should use corrected n values.

[2026-05-19] verify — molecules/compounds/tretinoin.md (R41 retinoid cluster anchor)

Pages verified: 1

  • molecules/compounds/tretinoin.md — verified: true (partial scope; Fisher 1996 cross-checked against R39-verified study page; identity fields confirmed via PubChem + ChEMBL APIs; Chien 2022 author list confirmed via Crossref; NCT06571721 confirmed RECRUITING; Siddiqui 2024 + Lin 2025 remain abstract-derived, no local PDFs)

Corrections applied (3):

  1. Lin 2025 adverse-event framing understated → CORRECTED: “more favorable adverse-event profile than tazarotene and glycolic acid” → “most favorable adverse-event profile overall; tazarotene and glycolic acid carried higher adverse-event risk” (abstract says “most favorable profile” unqualified)
  2. Lin 2025 hyperpigmentation superiority incomplete → CORRECTED: “tretinoin was superior for hyperpigmentation” → “tretinoin and retinol were superior for hyperpigmentation” (abstract names both)
  3. Lin 2025 gap tag wrong → CORRECTED: #gap/needs-replication (implies findings need replication) → #gap/no-fulltext-access (correct — PDF download pending; findings themselves are from published NMA)

Supersession check (R25): Three additional RCTs/systematic reviews found (PMIDs 40145274, 39921709, 38518171) — none supersede tretinoin core efficacy claims. PMID 39921709 (Kumari 2025 meta-analysis, n=337) found 0.1% tretinoin cream showed “lowest effect size” for hyperpigmentation among studied formulations — this contradicts the Lin 2025 “tretinoin superior for hyperpigmentation” finding and may warrant #gap/contradictory-evidence on the hyperpigmentation secondary efficacy claim in a future pass. Not a supersession of the core photoaging mechanism (Fisher 1996 is direct human in vivo data). literature-checked-through: 2026-05-19 confirmed.

[2026-05-19] verify — molecules/compounds/retinol.md (R41 retinoid cluster)

Pages verified: 1

  • molecules/compounds/retinol.md — verified: true (partial scope; Mambwe 2024 read in full from local PDF; all 9 cited DOIs confirmed via Crossref; identity fields confirmed via PubChem REST API)

Corrections applied (3):

  1. Potency ratio 5–20× not sourced to Mambwe 2024 → REFRAMED: the claim “5–20× less potent than tretinoin [^mambwe2024]” was unfounded — Mambwe 2024 nowhere states this numeric range. Reframed as widely-cited-but-unquantified with #gap/dose-response-unclear. The general claim of lower potency due to conversion steps is supported.
  2. ALDH1A2-predominance claim unsourced → REMOVED/FLAGGED: “ALDH1A2 predominates in human skin” was attributed to Mambwe 2024 but does not appear in the paper. Replaced with attribution-gap statement and #gap/needs-replication.
  3. Enzyme nomenclature in diagram and Step 1 tightened to Mambwe 2024 terms: diagram updated from “RALDH/esterase” entry to “retinyl ester hydrolase”; Step 2 enzyme named as “retinaldehyde dehydrogenase” per Mambwe 2024. Specific ADH isoforms (ADH1/4/7, RDH10) flagged as not from Mambwe 2024.
  4. Banner removed (auto-extraction banner deleted).

Supersession check: Lin 2025 NMA (PMID 40707570) is already the most recent synthesis and is cited on the page. No superseding post-2024 meta-analysis or large RCT found.

Downstream propagation needed:

  • interventions/pharmacological/retinoids.md — references retinol; if it states a specific potency ratio or attributes ALDH1A2 skin-predominance, those claims should be reviewed.
  • Any other page citing [^mambwe2024] for a specific potency ratio or ALDH1A isoform claim should be re-checked.

[2026-05-19] verify — interventions/pharmacological/retinoids.md (R41 class MOC)

Pages verified: 1

  • interventions/pharmacological/retinoids.md — verified: true (partial scope; Fisher 1996 + Sumita 2018 cross-checked against verified study pages; secondary sources not PDF-re-read)

Corrections applied (5):

  1. Fisher 1996 DOI WRONG → FIXED: 10.1038/nm1196-694 → 10.1038/379335a0 (seeder hallucinated a Nature Medicine DOI; correct journal is Nature 379:335-339)
  2. Fisher 1996 journal WRONG → FIXED: “Nature Medicine” → “Nature” in footnote
  3. Fisher 1996 n WRONG → FIXED: “n=26” → “n=6–17 (varies per experiment)” per verified study page (e.g., n=9 for AP-1 binding; n=7–10 for MMP endpoints)
  4. Fisher 1996 skin site WRONG in footnote → FIXED: “forearm” → “buttock skin in vivo” (the paper used buttock skin; forearm is from a different Fisher/Voorhees paper)
  5. Fisher 1996 archive status WRONG → FIXED: “no local PDF” → “local PDF confirmed in a local paper archive (download_status: completed)”
  6. Fisher 1996 AP-1 page reference MINOR: “pp.338-339” → “p.338” (text-stated on p.338 per verified study page)
  7. Sumita 2018 journal WRONG → FIXED: “Journal of Dermatological Treatment” → “J Eur Acad Dermatol Venereol” (correct journal for DOI 10.1111/jdv.15020)
  8. Sumita 2018 footnote expanded with full metadata (DOI, n breakdown, outcome summary, volume/pages, PMID) per verified study page
  9. RAR subtype specificity claim unsourced from Fisher 1996 → QUALIFIED: claim that “MMP suppression mediated via RAR-γ in keratinocytes and RAR-β in fibroblasts [^fisher1996]” is not in Fisher 1996; reframed with gap tag

Propagation needed:

  • Any other page with [^fisher1996] footnote citing doi:10.1038/nm1196-694 needs the same DOI correction

Supersession (R25): Lin 2025 NMA (PMID 40707570; already integrated on page) is the only recent large synthesis; no additional superseding RCT or meta-analysis found. literature-checked-through: 2026-05-19 confirmed.

Duplicate intervention-classes.md entry: lines 408 (topical retinoid / RAR agonist) and 1334 (rar-agonist) both exist. Line 408 is canonical (title-case values, broader name). Line 1334 is the duplicate. Propagation pass should delete line 1334 and ensure all 7 retinoid pages use title-case canonical values from line 408.

[2026-05-19] verify — molecules/compounds/bakuchiol.md (R41 retinoid cluster)

Pages verified: 1

  • molecules/compounds/bakuchiol.md — verified: true (partial scope; Chaudhuri 2014 and Puyana 2022 not_oa — abstract-level only; Dhaliwal 2019 bronze OA but Wiley paywall blocked local PDF download — verified via Crossref/PubMed abstract; Na Takuathung 2026 full PDF verified locally)

Corrections applied (5):

  1. ChEMBL ID wrong compound → CORRECTED: CHEMBL2364604 (buparlisib hydrochloride, a PI3K inhibitor) → CHEMBL262344 (bakuchiol). Critical identity error — wrong drug class entirely.
  2. CAS number missing → ADDED: null → 10309-37-2 (confirmed from PubChem CID 5468522 synonyms list). CAS 52662-39-2 found in some supplier catalogs is not canonical for this CID.
  3. CAS row missing from body identity table → ADDED: CAS 10309-37-2 row inserted into the Identity table; ChEMBL row corrected to CHEMBL262344.
  4. Gap tag needs-canonical-id for CAS → RESOLVED: Limitations section item 8 updated to document confirmed CAS 10309-37-2 and note CAS 52662-39-2 as a supplier-catalog variant to treat with caution.
  5. Footnote corrections: (a) Dhaliwal 2019 PMID 29947134 added; download status corrected from “pending” to “failed (Wiley paywall)”; (b) Fanning 2024 PMID 38564402 added; author order corrected (McGee SA was missing — now Fanning JE, McGee SA, Ibrahim OI); (c) Na Takuathung 2026 expanded with full quantitative results (MDs, CIs, p-values), full author list, article number, and genistein concentration (4%), and updated archive status to “PDF verified locally”; (d) Lau 2024 PMID 38758222 added with corrected full author list and Grade C recommendation detail.

Supersession check (R25): PubMed 2022–2026 search for bakuchiol+photoaging+skin aging (7 results) and bakuchiol systematic reviews/RCTs (5 results). No meta-analysis or large RCT (n>100) on bakuchiol monotherapy superseding Dhaliwal 2019 identified. Largest new RCT is Na Takuathung 2026 (n=80, combination formulation). literature-checked-through: 2026-05-19 confirmed.

Unverifiable claims: Chaudhuri 2014 (not_oa) — gene expression + clinical improvement figures remain abstract-level. Puyana 2022 (not_oa) — systematic review conclusions remain abstract-level. No quantitative claims from these sources were contradicted by available abstract data.

Downstream propagation needed: None identified. Bakuchiol page has no confirmed inbound links from entity pages citing the corrected ChEMBL/CAS identity fields.

Graph View

Backlinks