A Markovian growth dynamics on rooted binary trees evolving according to the Gompertz curve (Q453764)

The paper is inspired by cellular aging due to telomere shortening in mitosis (cell duplication) and studies a random mechanism able to produce the commonly observed (for instance in tumor growth, animal growth and regeneration) Gompertz curve. The assumption is that the process starts at time \(t=0\) with one founder cell at state (mitotic cycle) \(0\) and, at each mitotic cycle, the telomere shortens until it reaches, at exactly cycle number \(n\), a size that allows no further mitotic events. Each cell in state \(k\) (\(0\leq k\leq n-1\)), i.e.~one that results from exactly \(k\) mitotic events, will, independently from the others, duplicate (being replaced by two cells in state \(k+1\)) after a random time with exponential distribution with parameter \(\lambda_k=\beta (n-k)/n\) (\(\beta>0\)) related to its telomere length. Let \(X_n(k,t)\) be the number of cells at state \(k\) at time \(t\) and consider the Markov chain \(X_n(t)=(X_n(0,t),\dots, X_n(n,t))\), which will eventually stop changing when al cells are at state \(n\). Cell death is not considered. \\ This process can alternatively be seen as a ``growth Markov process taking values on the space of rooted binary trees, similar to the Aldous-Shields'' model [\textit{D. Aldous} and \textit{P. Shields}, Probab. Theory Relat. Fields 79, No. 4, 509--542 (1988; Zbl 0641.60026)]. The root is the founder cell and the cell population at time \(t\), \(Z_n(t)=\sum_{k=0}^n X_n(k,t)\), is formed by the active nodes (nodes with no descendants). The present paper shows that the expected value of \(Z_n(t)\) is \(S_n(t)=\left(1-\frac{e^{-\beta t/n}}{2}\right)^n\).\\ The paper looks at the limiting behaviour of the process for large telomere size, as \(n\rightarrow +\infty\). For that, it rescales the population size and time with \(\mathsf{X}_n(\tau)=2^{-n}Z_n(T_n+n\tau)\), where \(T_n=\frac{n}{\beta}\ln \left(\frac{n}{\ln 4}\right)+\frac{\ln 2}{2\beta}\). The expected value of \(\mathsf{X}_n(\tau)\) is \(W_n(\tau)=S_n(T_n+n\tau)\).\\ It is shown that \(\mathsf{X}_n(\tau)\) converges in distribution to a r.v.~equal to the (Gompertz curve) value \(\exp \left(-(\ln 2) e^{-\beta \tau}\right)\) with probability one and that \(W_n(\tau)\) also converges to the Gompertz curve. Using a theorem in [\textit{J. Jacod} and \textit{A. N. Shiryaev}, Limit theorems for stochastic processes. Berlin etc.: Springer-Verlag (1987; Zbl 0635.60021)], this paper proves a central limit theorem: the martingale \(M_n(\tau)=2^{n/2}\left(\mathsf{X}_n(\tau)-\mathsf{X}_n(0)-\frac{n}{2^n}\sum_{k=0}^{n}\lambda_k\int_{0}^{\tau}X(k,T_n+ns)ds\right)\) converges in distribution to a Brownian motion \(B_{v(\tau)}\), with \(v(\tau)=\exp \left(-(\ln 2) e^{-\beta \tau}\right)-1/2\). Some consequences on the average telomere length are then deduced.