context: dispatched wiki-verifier on Hickson 2019 (doi:10.1016/j.ebiom.2019.08.069) to settle whether the adipose p16/p21/SA-βgal reductions are % positive cells or mRNA — the distinction determines whether the D+Q model kill-fraction ε_kill (model/dq-calibration-design.md) is a true per-cell kill. PDF read end-to-end (Results, Methods, Figs 1–5).
RESOLVED: all three are % positive CELLS (IHC counts per 100 adipocytes; SA-βgal⁺ cells per 100 nuclei), NOT mRNA. So cellular-senescence.md (which said “mRNA”) was wrong; the compound/intervention pages were right. ε_kill is a legitimate per-tissue kill fraction (caveats: IHC not perfectly senescence-specific; per-adipocyte denominator could be confounded by stromal:adipocyte ratio).
2 corrections surfaced:
cellular-senescence.md: “p16 mRNA −35%, p21 mRNA −17%” → ”% positive cells (IHC, not mRNA)”; added skin values — the wiki’s “directionally concordant but unquantified” was wrong; skin was fully quantified (p16 −20% p=0.026, p21 −31% p=0.016, Fig. 4).
dasatinib.md: “+86%” adipocyte-progenitor replicative potential → “+8%” — a digit-transposition that survived the prior (2026-05-08) verification round (paper: +8%, n=11, p=0.027, MTS passage-3). The feedback_verifier_introduced_error pattern.
updated (verifier): cellular-senescence.md, dasatinib.md, + verified-scope on all four (cellular-senescence / dasatinib / quercetin / senolytics); quercetin.md + senolytics.md needed no factual fix (already “cells” / “+8%”).
sim fallout (model/, separate commit): dq-calibration-design.md — cells-vs-mRNA marked RESOLVED; +86%→+8% fixed (2 spots); skin folded in as a 2nd-tissue anchor (ε now bracketed by two tissues; IHC markers cluster 0.17–0.35); cleared #gap/hickson-cells-vs-mrna-labeling + #gap/no-second-tissue-human-anchor; residual #gap/ihc-not-perfectly-senescence-specific. Also added an engine-run HOLD verification (Step 5) the prior day’s work had deferred.
context: user flagged a new Cell senescence paper (Suryadevara et al. 2026, NIH SenNet Perspective), then asked to branch out to companion data papers “if the data improves the wiki.”
propagated:hallmarks/cellular-senescence.md (new ## Senescent-cell heterogeneity and the senotype concept + empirical-support block citing both preprints); processes/sasp.md (senotype cross-ref); interventions/pharmacological/senolytics.md (new ### Senolytic resistance and heterogeneity — csV1B2 → BH3-mimetic resistance biomarker).
main-agent corrections of seeder output: fixed a fabricated section anchor ([[sasp]] § "Cell-type-specific SASP variants" → § "SASP composition"; real content = astrocyte-complement/fibroblast-MMP SASP; cardiomyocyte non-canonical SASP lives on cardiac-fibrosis.md); corrected SenLect description (genetically-encoded purification system, not an AI tool).
not seeded (future leads): SenLect (Sargeant/Carosi method preprint, non-SenNet); a definitive published SenNet Cell atlas DATA paper could not be located (only the Perspective is indexed) and the press-release blood biomarkers (CKD/frailty/diabetes) belong to companion data papers not yet locatable — gap/future-seeding.
model relevance (deferred per user): tissue-specific/heterogeneous senescence + an apoptosis-resistant (senolytic-evading) subset + the immune-clearance-decline framing bear directly on the lumped-senescence-node and the (frozen) clearance-capacity design.
leak-gate: clean (3 checks). All attribution accurate (Krizhanovsky preprints ≠ NIH SenNet).
follow-up (same day): two genuine SenNet companion DATA papers located + seeded (user pointed to the Cell SenNet consortium page). Updates the “could not be located” note above:
studies/anerillas-2026-sencat.md — SenCat (Anerillas et al., Molecular Cell 2026, doi:10.1016/j.molcel.2026.05.017, PMID 42276073; peer-reviewed SenNet/NIA; paywalled, abstract-sourced, verified:false). Multi-omic (transcriptome+proteome) profiling of 14 primary human cell types × >30 senescence paradigms: no single universal senescence marker (catalog-scale support for senotype heterogeneity), shared metabolic + damage-response/repair pathways, ML SenCat signatures for cross-dataset senescence scoring. Propagated to cellular-senescence.md senotype section (now the strongest peer-reviewed support there).
studies/farzad-2026-immunosenescence-atlas.md — spatial multi-omics immunosenescence atlas of human lymph nodes; germinal-center B-cell alteration (Farzad et al., Cell Press Blue 2026, doi:10.1016/j.cpblue.2026.100053; SenNet; abstract NOT retrievable — title/press-only, verified:false, gap/abstract-unavailable). Propagated one flagged line to immunosenescence.md § B-cell compartment.
Updated the cellular-senescence.md scope caveat (these two are now seeded; remaining leads = broader cross-tissue atlas + blood biomarkers CKD/frailty/diabetes + SenLect).
leak-gate: clean.
follow-up (same day): seeded the foundational SenNet detection consensus + saved the biomarker resource (user pointed to docs.sennetconsortium.org/biomarkers + the backing Google Sheet):
studies/suryadevara-2024-sennet-detection-recommendations.md — “SenNet recommendations for detecting senescent cells in different tissues” (Suryadevara et al., Nat Rev Mol Cell Biol 2024;25(12):1001–1023, doi:10.1038/s41580-024-00738-8, PMID 38831121; 79 authors; green-OA, abstract-sourced, verified:false). The canonical per-tissue senescence-detection consensus (14 mouse+human tissues; molecular signatures + morphology + circulating markers; no single universal marker) — predecessor to the 2026 senotype Perspective (same first author). Was missing from the wiki.
sources/sennet-biomarkers.csv (+ sources/sennet-biomarkers.md provenance) — snapshot of the SenNet biomarker Google Sheet: 703 curated biomarker entries × 11 columns (marker, HGNC_ID, biomolecule, tissue, cell, context, organism, hallmark, PMID citation, assay) across the 14 tissues incl. “Circulating markers”. Kept as a discovery/reference index (cite the per-row PMIDs, not the snapshot); living doc, re-pull for updates.
propagated: cellular-senescence.md § Definition — SenNet-2024 detection note (tissue-specific multi-marker panels) + footnote + link to the biomarker snapshot.
remaining SenNet leads: a broader cross-tissue single-cell/spatial atlas data paper + the specific 2026 blood-biomarker predictors (CKD/frailty/diabetes) still not pinned to retrievable DOIs (the 703-row resource is the curated marker index, not the 2026 predictive-biomarker paper). gap/future-seeding
context: user flagged the bioRxiv preprint “Selective Elimination of TP53 Mutant Cells by Transcript-Activated Chromatin Shredding” (doi:10.64898/2026.05.08.723607) as an interesting CRISPR cell-targeting method. Recency check found it is now peer-reviewed in Nature (doi:10.1038/s41586-026-10738-7, “Targeting Cancer-Specific Mutations with RNA-Triggered Chromatin Shredding”, 2026-06-08), with a companion foundational paper (Scholz 2026, Nature, doi:10.1038/s41586-026-10466-y) and the origin paper (Dmytrenko 2023, Nature, doi:10.1038/s41586-022-05559-3). Cited the Nature version as primary; verified against the openly-licensed (CC-BY) preprint full text.
added (3 pages):
studies/zeng-2026-cas12a2-chromatin-shredding.md — primary study; verified:true (claude, against CC-BY preprint; Nature version paywalled/unread). SuCas12a2 trans/collateral cleavage → chromatin shredding → DNA-damage death; selectivity via SNV (R248Q G→A creates the “mutA” PFS) / indel-junction (EGFR E746_A750del) / oncogene-abundance (CCNE1/MYC); co-culture ratio 0.62→4.5; 25.7% of TP53 muts B→A, 73.2% PFS-usable (n=16,708); LNP mRNA in vivo reduced MYC-liver-tumour surface area + delayed PC9-lung metastasis but did not clear advanced burden (~7–18% delivery).
methods/cas12a2-trans-cleavage.md — the technique (verified:false; Zeng-attributed claims verified, Dmytrenko/Scholz metadata-only gap/no-fulltext-access). Framed as the opposite design point to in-vivo-base-editing (maximal non-specific damage gated by an RNA sensor vs precision DSB-free edit).
interventions/gene-therapy/transcript-activated-cell-ablation.md — the modality page (mode:gene-therapy, preclinical, human-evidence preclinical-only). Carries the 4-modality comparison table (BH3-mimetic / CAR-T / suicide-gene / transcript-activated) and the honest aging framing.
aging relevance (explicitly hypothesis, not result): a non-SCAP, non-apoptosis-priming programmable kill switch — candidate for senotype-targeted senolysis — but bounded by the SenNet no-universal-marker problem (no senescence-exclusive trigger transcript; p16/CDKN2A not specific) and by delivery. Plus a p53-aging cameo: authors kill mutant-p53 cells rather than activate p53 (activation → senescence/whole-genome-duplication; the Tyner-2002 tradeoff).
propagated (5 pages): senolytics.md (§ Related approaches — new modality bullet), cellular-senescence.md (§ senotype concept — transcript-addressed-killing paragraph), sens-damage-categories.md (ApoptoSENS modality list + related pages), phenotypes/cancer.md (§ Modern therapeutic landscape bullet + footnote), molecules/proteins/p53.md (§ Pharmacology — ablation-vs-reactivation bullet + footnote).
verification: dispatched wiki-verifier against the preprint full text — all quantitative claims confirmed, zero corrections; flipped study page verified:true.
not seeded (DOI-cited leads): Scholz 2026 (companion eukaryotic-killing) + Dmytrenko 2023 (origin) — candidates for their own study pages if the platform earns deeper coverage. gap/future-seeding
leak-gate: clean (checks run against the 3 new files + the 5 modified pages + this log).
[2026-06-13] ingest | autonomic / cholinergic anti-inflammatory axis — 3 new pages (HRV biomarker, CAP pathway, α7nAChR)
context: user asked for an assessment of a perspective paper (Errico et al., “Sympathetic-parasympathetic system deregulation theory of aging”, npj Aging 2025, doi:10.1038/s41514-025-00293-2) surfaced via a gethealthspan.com article. Assessed: it is a theory/perspective with no new data, in npj Aging (Nature Portfolio, NOT flagship Nature), with a material COI — lead author is an electroCore (VNS device) consultant + Vagus Nerve Society board member, and the work was funded by the Vagus Nerve Society, whose conclusion (VNS = anti-aging) directly benefits the funder. Defensible core (cholinergic anti-inflammatory reflex, HRV as mortality biomarker, sympathetic-overdrive effects) vs overreach (autonomic imbalance as THE unifying upstream driver of all hallmarks; “VNS extends lifespan” untested). User decision: do NOT reference the Errico paper anywhere — seed the genuinely-missing underlying entities from canonical primary literature instead.
added (3 pages, all verified:true after adversarial verifier pass):
molecules/proteins/chrna7.md — type: protein (α7nAChR). IDs verified against live DBs (UniProt P36544, NCBI 1139, HGNC:1960, Ensembl ENSG00000175344, mouse Chrna7). Wang 2003 essential-receptor anchor; CHRFAM7A human-specific dominant-negative (15q13.2); pharmacology (GTS-21, encenicline Phase-3 GI-safety hold); mr-causal-evidence not-tested; druggability-tier 2.
verifier corrections (adversarial pass earned its keep): HRV page — Geurts 2023 MR claim was sign-inverted (page said MR did NOT support HRV→AF; paper shows it DID, OR ~1.6) → fixed; Tsuji 1994 n 2,501→736 (substudy, not full Framingham); Dekker case-cohort design clarified; El-Malahi scope narrowed to CHF subgroup; Tegegne n’s split out. CAP page — Rosas-Ballina ChAT⁺ frequency 1–2%→4.4%, host strain Rag2⁻/⁻→nude, Peña JAK2–STAT3 direction corrected (prevents STAT3 phosphorylation, not “activates JAK2”), Koopman IL-1β attributed to epilepsy sub-cohort not RA. CHRNA7 page — CHRFAM7A location 15q13.1→15q13.2, Courties “multi-joint”→right-knee-only + human-osteoblast null added.
downstream-correction grep (verifier-flagged): checked geurts / tsuji / rosas-ballina / peña2010 / 15q13.1 across the wiki — all hits false positives (Geurts L = Everard 2013 co-author; “2,501” unrelated; “rag2” = RAG2 gene; “rosas” = Rosas-Campos 2025). No other page carried the wrong versions.
discipline note: every propagated autonomic→aging causal pointer explicitly hedged as mechanistic/associational, not a proven causal driver — deliberately avoiding the Errico overreach the user rejected.
leak-gate: clean (3 checks, against new + modified files + this log).