saturating-removal-model

⚠️ Auto-extracted by Claude on 2026-06-24 — synthesis MOC drafted from the full PDF of Raz et al. 2026 (10.1038/s43587-026-01138-7) plus the originating Karin & Alon 2019 model; the 2019 paper’s claims are attributed from training knowledge + the 2026 paper’s framing, not independently re-verified against the 2019 PDF. Verify quantitative claims before relying on them.

Saturating Removal (SR) model — damage accumulation and the two aging regimes

Mode B conceptual frame (per writing-hypothesis-pages): a mechanistic mathematical model of how organismal damage accumulates and produces the empirical laws of mortality, rather than a sharply falsifiable biological hypothesis. It organizes the demographic facts of aging (Gompertz / Weibull hazard curves, lifespan scaling across species) under a single low-dimensional dynamical model.

The SR model proposes that aging can be summarized by a single scalar damage variable x whose dynamics are governed by production, saturating removal, and noise, with death as a first-passage event when x crosses a threshold. It was introduced by Karin & Alon (2019) to explain why senescent-cell turnover slowing with age yields the Gompertz law ¹, and extended comparatively across nine species by Raz et al. (2026) ².

---

The core idea

A 1-D stochastic differential equation captures damage dynamics ¹²:

$\frac{dx}{dt}=\underset{\text{production}}{\underbrace{\eta t}}-\underset{\text{saturating removal}}{\underbrace{\beta \frac{x}{\kappa +x}}}+\underset{\text{noise}}{\underbrace{\sqrt{2\epsilon }\xi }}$

• Production rises linearly with age (ηt). This linear rise is required to reproduce the Gompertz law (exponentially rising hazard); biologically it reflects accumulation of damage-producing units (e.g., senescent cells produced at a constant per-unit rate that itself grows).
• Removal saturates at high damage (Michaelis–Menten form), a generic property of biological clearance — e.g., immune removal of senescent/damaged cells. If removal did not saturate, steady-state damage would be flat and aging would not accelerate.
• Noise ε is biological stochasticity; death is the first passage of x across threshold X_c.
• A Makeham term m_ex adds extrinsic (age-independent) mortality.

The model distinguishes removal (clearing existing damage — the immune system) from repair (lowering the production parameter η — e.g., DNA repair preventing damage from forming). See raz-2026-sr-model-aging-regimes for the full parameter table.

---

What the frame predicts / organizes

1. The two aging regimes

The ratio of production (ηt) to removal (β) splits species into two qualitative regimes ²:

	Ballistic aging	Quasi-steady-state aging
Dynamics	production outpaces removal; damage rises like a free-falling object ⟨x⟩~½ηt²	production balanced by removal; damage tracks a slowly rising set-point
Production–lifespan scaling	η ~ 1/L²	η ~ 1/L
Hazard shape	Weibull-like (power-law)	Gompertz-like (exponential, slowing at very old age)
Species	yeast, C. elegans, most flies, mice	humans, dogs, cats, guinea pigs

Organisms lie on a continuum; some (e.g., female mice, starving E. coli, Drosophila strain 853) straddle the transition.

2. Production rate η is the master lifespan knob

Across nine species η spans ~7 orders of magnitude and is the best single predictor of lifespan (SHAP importance 2.85, vs 1.53 / 1.38 / 0.33 for threshold / removal / noise). Removal, noise, and threshold are nearly invariant across mammals. Evolution appears to set lifespan chiefly by tuning damage production (molecular accuracy), not removal or robustness ².

3. Conserved mammalian parameter combinations

• Noise→threshold timescale T_c = X_c²/ε ≈ 600 days (factor ~3) across mice, guinea pigs, cats, humans.
• Removal/noise timescale ratio βX_c/ε ≈ 10 (factor ~2).
• Death threshold X_c ≥ ~10× κ (removal saturates in youth) in every organism.

These near-universal ratios hint at conserved removal/noise machinery — one hypothesis being circadian fluctuation in immune clearance as the source of biological noise.

---

Relationship to other frames

• Mortality laws. The SR model gives a mechanistic basis for the otherwise purely descriptive Gompertz and Weibull hazard functions — previous fits (Gompertz, Weibull, logistic) had no mechanistic grounding. Gompertz ⇔ quasi-steady-state; Weibull ⇔ ballistic.
• Damage-accumulation theories. It is the quantitative, demographic-scale companion to the molecular damage theories — somatic-mutation-theory-of-aging, dna-damage-theory-of-aging — which supply candidate production mechanisms (the inverse relation between somatic-mutation / translation-error rates and mammalian lifespan).
• Cellular senescence. Senescent cells are the leading candidate for the damage variable x in mammals; the model was originally calibrated on senescent-cell turnover ¹.
• Model-organism extrapolation. Regime classification gives a principled answer to which organisms model human aging — see _extrapolation-guide.

---

Intervention implications

• Target production (η) for lifespan extension — it is the parameter evolution actually varies. Removal-targeting (e.g., senolytics) and threshold are lower-leverage for maximal extension.
• But production-only interventions stretch sickspan as well as lifespan. To extend life and compress morbidity, an intervention must also raise the threshold X_c, reduce noise ε, or increase removal β ².

---

Status & what would update this frame

Active-frame. As a Mode B frame the SR model is judged by descriptive adequacy and fruitfulness, not by a single falsifying experiment. It fits survival curves across four orders of magnitude of lifespan well; it currently misses the late-age mortality plateau in C. elegans. It would be strengthened by: independent biological measurement of x (e.g., direct senescent-cell burden trajectories matching the inferred η/β), and challenged by: organisms whose hazard shape cannot be captured, or evidence that lifespan differences track removal/threshold rather than production. Because the parameters are inferred from the same survival data they explain, they are mathematically grounded hypotheses awaiting experimental verification, not measured biological quantities.

---

Footnotes

Footnotes

1. karin-2019-senescent-cell-turnover-gompertz · Karin O, Agrawal A, Porat Z, Krizhanovsky V, Alon U. “Senescent cell turnover slows with age providing an explanation for the Gompertz law.” Nat Commun 10:5495 (2019) · doi:10.1038/s41467-019-13192-4 · introduces the Saturating Removal model; calibrated on senescent-cell turnover in mice · model: mouse. ↩ ↩² ↩³

2. raz-2026-sr-model-aging-regimes · Raz N et al. · Nat Aging 6:1330–1340 (2026) · doi:10.1038/s43587-026-01138-7 · comparative SR fit across nine species; η best lifespan predictor; ballistic vs quasi-steady-state regimes · design: in-silico modeling of observational mortality data. ↩ ↩² ↩³ ↩⁴ ↩⁵