Log — 2026-06-24

[2026-06-24] ingest | Raz et al. 2026 Nature Aging — Saturating Removal damage-accumulation model + the two aging regimes

  • context: user supplied a new open-access paper (DOI 10.1038/s43587-026-01138-7, Nat Aging 6:1330–1340, Uri Alon lab) and asked to incorporate it. It is the comparative-across-species application of the SR model (origin: Karin & Alon 2019, Nat Commun 10:5495). The wiki had no SR-model page and no mathematical mortality-dynamics frame at all (Gompertz/Weibull were only mentioned in passing on negligible-senescence) — a genuine gap.
  • page-type decisions: split into a study page (primary extraction) + a Mode B hypothesis page (the model itself as a conceptual frame), per the data-vs-synthesis division precedent. The model is a quantitative frame, not a sharply falsifiable bio-hypothesis → Mode B (treatment-mode: conceptual-frame, status: active-frame).
  • added (2 pages):
    • studies/raz-2026-sr-model-aging-regimes.md — full multi-species extraction: SR equation (dx/dt = ηt − βx/(κ+x) + √(2ε)ξ, death = first passage of x past X_c); 9-species dataset table (yeast/worm/fly/E.coli/mouse/guinea-pig/cat/dog/human, n 674–80k); η best lifespan predictor spanning 7 orders of magnitude (SHAP log(η)=2.85 vs X_c 1.53 / β 1.38 / ε 0.33); ballistic (Weibull, η1/L²: yeast/worm/fly/mice) vs quasi-steady-state (Gompertz, η~1/L: human/dog/cat/guinea-pig) regimes; conserved mammal params T_c=X_c²/ε≈600d and βX_c/ε≈10 and X_c≥10κ; senescent cells = candidate damage substrate; intervention implication (target production η, but production-only stretches sickspan → also need threshold/noise). verified: false (see verification note below).
    • hypotheses/saturating-removal-model.md — Mode B frame; aliases [SR model, ballistic aging, quasi-steady-state aging, two aging regimes]; aggregates Karin 2019 (origin) + Raz 2026 (comparative); maps the regimes onto Gompertz/Weibull; intervention lever-hierarchy; “what would update this frame.” verified: false (depends partly on the unread Karin 2019 PDF; its footnote tagged gap/needs-verification).
  • propagated (4 existing pages):
    • model-organisms/_extrapolation-guide.md — new “Mortality-dynamics lens” section: regime classification as a demographic axis orthogonal to pathway/phenotype conservation. Key caveat surfaced: mice are in the ballistic regime (male further from human-like than female) → extra reason to distrust mouse lifespan-curve extrapolation; dogs/cats/guinea-pigs demographically closest to humans.
    • hallmarks/cellular-senescence.md — § “Two arms of intervention: production rate and clearance” extended: this production-vs-clearance split IS the SR model’s biological substrate; the demographic analysis says production (η) is the lifespan knob and removal (β) is near-invariant across mammals. Added [^raz2026] footnote.
    • hypotheses/negligible-senescence.md — Related-hypotheses bullet: SR model supplies the mechanistic basis for the Gompertz/Weibull fits used to test (non-)senescence; a genuinely negligibly-senescent species sits outside both regimes (η→0). NMR is the untested candidate.
  • verification posture (deliberate): I read the full open-access PDF end-to-end (all 12 pp incl. methods + Extended Data Table 1) and ran a second-pass numeric re-check against the source text + figures, which caught one extraction error — the removal/noise timescale ratio is βX_c/ε≈10, NOT βκ/ε (corrected on both pages). I also softened the yeast β=0 asymptotic-hazard claim to its load-bearing form (h(t) ∝ t²) rather than assert an uncertain prefactor exponent. BUT no independent wiki-verifier adversarial pass ran: the paper is not in the local archive (download start → “DOI not found in database”) and is too new for PMC. Per the wiki’s two-layer epistemics I did not self-certify — both pages stay verified: false with banners stating it’s a high-fidelity single-agent extraction with an independent pass pending. Flip when the archive ingests the DOI (or on user’s OK for single-agent verification).
  • recency: study is brand-new (pub. 2026-06-09); hypothesis literature-checked-through: 2026-06-24.
  • leak-gate: clean (checks #1 + #3 on new/changed .md; the Downloads PDF path was used only for my own reading and appears in no tracked file).
  • index.md: no change (folder-level catalog, not per-page).

[2026-06-24] ingest+verify | Karin & Alon 2019 Nat Commun — origin paper of the Saturating Removal model (the mechanistic senescence anchor)

  • context: follow-on to the Raz 2026 ingest — user asked to seed+verify the SR model’s origin paper, which the two new SR pages footnote as [^karin2019] and which had no dedicated page. It’s the empirical (mouse) calibration behind the abstract cross-species model.
  • DOI correction (memory: seeder-brief DOI unreliable): my recalled DOI 10.1038/s41467-019-13192-3 404s on Crossref; the correct DOI is -4 (10.1038/s41467-019-13192-4), title “Senescent cell turnover slows with age…” (no “down”), Nat Commun 10:5495. Fixed the wrong -3 DOI in BOTH already-created SR pages (raz study + saturating-removal-model hypothesis) and pointed their [^karin2019] footnotes at the new study page.
  • added (1 page, seeded → adversarially verified → verified:true):
    • studies/karin-2019-senescent-cell-turnover-gompertz.md — PMID 31792199, PMC6889273 (OA). Verified against the PMC open-access full text (the paper was not yet retrievable locally, so the PMC copy was used).
  • verifier corrections (the adversarial pass earned its keep — 12 fixes, two conceptually important):
    1. Production term: the seeder wrote that the 2019 model uses constant production α and that linear-rise ηt is a 2026 addition — WRONG. ηt originates in Karin 2019. Corrected equation to dX/dt = ηt − βκX/(κ+X) + ε dW. ML params added (η=0.15 day⁻¹yr⁻¹, β=0.27 day⁻¹, κ=1.1, ε=0.14, X_C=17).
    2. Removal mechanism (the big one): seeder attributed the age-related removal slowdown to extrinsic NK-cell/immune aging — WRONG. The paper’s final SR model uses mechanism (iv): density-dependent self-inhibition — per-cell removal rate β/(κ+X) falls as SnC abundance X rises (saturation), a positive feedback / “critical slowing down.” Extrinsic immune aging (mechanism iii) was tested and NOT required to fit the data. Corrected throughout.
    3. Exact half-lives: ~5±1 d (3 mo) → ~25±6 d (22 mo), ~5× residence-time increase (seeder’s “days→weeks / ~3d→~3wk” was a guess). Method: p16-luciferase TBL imaging, n=33 longitudinal + bleomycin pulse-chase (1.5 U/kg intratracheal); C57BL/6; n-subjects=82. Lung-localized induction flagged as a limitation.
  • propagation (main agent): folded the Karin mechanism into hallmarks/cellular-senescence.md § “Two arms” — sharpened a conflation the verifier flagged: the SR removal-slowdown is abundance-driven saturation (intrinsic, density-dependent), mechanistically distinct from the age-related immune-clearance decline documented elsewhere on the page (immunosenescence). Karin found immune aging is not required to explain the turnover slowdown. Added [^karin2019sr] footnote. (This also corrects a slight mischaracterization I’d given the user verbally, who I’m flagging it to.)
  • link hygiene: fixed broken [[processes/cellular-senescence]] (canonical is hallmarks/) and aligned [[molecules/proteins/p16]][[p16]] (dominant dangling form; no p16 page yet — legitimate stub-target) on the new page.
  • supersession: 2022–26 search found nothing revising the turnover-slows-with-age claim; Raz 2026 (already in wiki) is the extension.
  • leak-gate: clean.

[2026-06-24] sim | SR saturating-removal integration IMPLEMENTED (Path B) + Codex-reviewed

  • context: follow-on to the Raz/Karin ingest — folded the Saturating Removal model into the simulator’s senescence/clearance/senolytic machinery. Design + calibration in model/sr-saturating-removal-integration-design.md (§11, two Codex GPT-5.5 rounds); user chose Path B (batch the whole change with best-guess evidence-backed values).
  • changed (hand-edited logic): model/engine.mjs (clearance fold → SR branch: dx/dt = coeff·(S−T) − [R(S;VmaxEff) − R(T;Vmax)], R(S)=Vmax·S/(Km+S), deviation-form, production-response self-amp activated, immunosenescence→Vmax modifier, pulse-separation guard preserved); frameworks/causal-graph-parameters.md (cellular-senescence clearance block: form=saturating, Vmax=0.075, Km=0.30, betaImm=0.010, productionResponse=true — the source-of-truth json block) + D+Q preset rebound 1/3/8→0.25/0.5/1.5 yr; model/test.mjs (SR test block rewrite + re-baselines).
  • regenerated artifacts: model/params.json (via build-params.mjs from the markdown), viz/aging-simulator.html (via build-app.mjs — embeds engine+params).
  • invariants HELD: baseline LE byte-identical M 77.45885545055621 / F 82.11784965025268, max|B−T|=0; 322/322 tests pass. No-age-pegging preserved (removal reads state S, production stays self-amp not ηt).
  • calibration (best-guess anchored, §11): Vmax/Km tip the deviation system into loss-of-resilience at age ≈93 (human quasi-steady-state, Raz 2026); structure from Karin 2019 (mechanism iv). D+Q ΔLE@55 re-baselined 0.01865→0.0053 (shorter 0.5-yr rebound ⇒ more transient senolytic, clinically realistic).
  • bug caught + fixed (testing earned its keep): initially used VmaxEff for BOTH R(S) and R(T) ⇒ immunosenescence/restoration cancel at S=T; fixed so only R(S) uses the live ceiling.
  • Codex code-review (GPT-5.5 high, read-only): ran the suite itself (322/322), confirmed baseline-invariance/sign/pulse-separation/no-double-count + the re-baselines; flagged 2 (legacy c0 double-heal test needed form:linear; prodResp clamp-consistency) — both fixed.
  • NOTE: loop activation (reverse infl→sen) remains STUBBED; compound ε_kill wiring beyond the D+Q exploratory preset still deferred. Not committed (awaiting user OK).

[2026-06-24] ingest | 2 natural senolytic compounds — procyanidin-c1 + piperlongumine (ad-hoc; D+Q-alternative context)

  • context: user asked for OTC-available senolytic alternatives to D+Q (dasatinib is prescription-only). Seeded+verified the two missing natural senolytics flagged in the senolytic comparison. Ad-hoc user request → dispatched directly (not via ROADMAP).
  • added (2 compound pages, parallel seed → parallel background-verify, both verified:true):
    • molecules/compounds/procyanidin-c1.md — grape-seed B-type procyanidin trimer (PubChem 169853, ChEMBL CHEMBL290632, CAS 37064-30-5, C45H38O18). Landmark: Xu 2021 Nat Metab (doi:10.1038/s42255-021-00491-8, PMC8688144, full-PDF verified). Dual dose-dependent: senomorphic (SASP-suppress) low / senolytic ≥50 µM. Mechanism: NOXA/PUMA↑→ROS→Δψm collapse→caspase 3/7 + p53 mito-translocation. Aged C57BL/6J (24–27 mo, 20 mg/kg biweekly): whole-life median +9.4% (843→922 d), post-treatment median +64.2% (123→202 d), HR 0.35. Seeder added senomorphic + ros-induction mechanism classes to intervention-classes.md.
    • molecules/compounds/piperlongumine.md — long-pepper (Piper longum) alkaloid (PubChem 637858, ChEMBL CHEMBL465843, CAS 20069-09-4, C17H19NO5). Landmark: Wang 2016 Aging (doi:10.18632/aging.101100, full-PDF verified) + Zhang 2018 OXR1 (Aging Cell). Mechanism: targets OXR1 → chronic antioxidant-defense erosion → ROS-dependent senolysis (EC50 ~20 µM normal vs ~6–8 µM senescent, ~2.5–3.3× window); navitoclax-synergistic (CDI 0.02–0.41). Preclinical, 0 trials.
  • verifier corrections (both adversarial passes earned their keep):
    • PCC1: removed a FALSE BCL-2/BCL-xL targets claim (ChEMBL CHEMBL290632 has 0 binding entries vs BCL-2/BCL-xL; PCC1 actually lowers BCL-2) → targets:[], mechanism rewritten to NOXA/PUMA. Fixed a misattributed PMID (39878685 = Gan 2025 FASEB renal, NOT Liu 2025 NPJ Aging) + a wrong seeder-guessed Liu 2024 PNAS DOI (2409734121→2311028121); CAS resolved; NCT06641869 sponsor → Express Rx Inc (industry), no results posted; SASP factor list pinned (IL-6/CXCL8/MMP3/SPINK1/WNT16B/GM-CSF/IL-1A/AREG).
    • PL: adjudicated the seeder-flagged mechanism discordance — Wang 2016 “ROS-independent” (acute ROS scavenging via GT3 doesn’t rescue SC death) + Zhang 2018 “OXR1→chronic-ROS” are COMPLEMENTARY (OXR1-upstream; acute scavenging misses the chronic antioxidant-erosion kill), not contradictory → #gap/contradictory-evidence removed. Pinned EC50/selectivity + ABT-263 CDI synergy; corrected 4 wrong citation fields (Liu 2018 issue/pages, Zhang 2024 volume, Dotou 2024 pages); kept Ali 2025 negative bone-loss result framed honestly.
  • propagation (main agent): added both to interventions/pharmacological/senolytics.md — the compound list + the mechanism table (linked the pre-existing unlinked PCC1 row + corrected its mechanism away from “less-studied” to the verified NOXA/PUMA dual-mode; added a PL row). [[oxr1]] is a stub-target (no page); no stale BCL-2/xL PCC1 claim elsewhere (both compounds were new).
  • simulator context: these slot into the senolytic-pulse comparison (fisetin ε_kill ~0.15 / quercetin-mono ~0.08 / D+Q ~0.35); PCC1/PL not yet assigned ε_kill or wired as presets.
  • leak-gate: clean (checks #1+#3 on new/changed; new files grepped explicitly).

[2026-06-24] sim | wired 4 senolytic UI presets (fisetin/quercetin-mono/PCC1/piperlongumine) + flagged the lumped-S combination limitation

  • added to frameworks/causal-graph-parameters.md operatorPresets (source → build-params → params.json → build-app → viz): fisetin-one-off (ε_kill 0.05/0.15/0.30), quercetin-mono-one-off (0.03/0.08/0.15), pcc1-one-off (0.05/0.12/0.25), piperlongumine-one-off (0.03/0.10/0.20); all rebound 0.25/0.5/1.5 yr. ε_kill = §11 best-guesses (fisetin anchored; quercetin/PCC1/PL placeholder, PCC1/PL weakest — preclinical-only). Each provenance carries the SCAP cell-type target + the combination caveat.
  • 322/322 unchanged (presets inert until selected); baseline byte-identical.
  • Combination limitation documented (user-flagged): the SINGLE lumped senescence node cannot represent SCAP cell-type selectivity. Co-active senolytic pulses ADD on one pool — demo: yearly singles ΔLE fisetin 0.147 / quercetin 0.085 / PCC1 0.121 / PL 0.103; all-four-stacked → effective kill ~0.45/pulse, sen@80 0.251→0.160, ΔLE 0.331. Biologically the four substantially OVERLAP (quercetin’s endothelial pool ⊂ fisetin; PL’s broad ROS overlaps all), so additive stacking OVER-credits; the model also can’t represent the macrophage/DC pool fisetin spares (a floor). Proper fix = cell-type sub-pool disaggregation (clearance-state-design.md Q6, deferred). Even the over-credited stack only buys 0.33 yr (downstream dilution dominates).
  • not committed.