Diagnosing Overshoot in Maturation Models

Why a Mature-Cell Peak Can Lag Behind an Input Change

A rise in precursor input does not necessarily produce an immediate rise in mature-cell count. Precursors must mature before they contribute to the mature compartment, while mature cells continue to be lost throughout the response. As a result, the mature-cell count can keep rising for a time after precursor input has begun to fall. Diagnosing this overshoot requires separating the flux into the mature compartment from the loss out of it.

The earlier material on three-stage maturation models and homeostatic cell replacement provides the relevant compartment perspective. Here, the focus is a balance-based diagnosis: what trajectories and fluxes would support delayed maturation, and what evidence would instead indicate a change in mature-cell loss?

Definition: Let $P(t)$ and $M(t)$ denote precursor and mature-cell counts. Let $F(t)$ be the maturation flux into the mature compartment, measured in cells per unit time, and let $L(t)$ be the mature-cell loss flux in the same units. The mature-cell balance is $$ \frac{dM}{dt}=F(t)-L(t). $$ Under first-order mature-cell loss with a constant rate $k_m>0$, $L(t)=k_mM(t)$, so $$ \frac{dM}{dt}=F(t)-k_mM(t). $$

In the simple two-compartment model used here, precursor cells mature at a constant rate $k_p>0$, and input $u(t)$ enters the precursor compartment:

\frac{dP}{dt}=u(t)-k_pP(t), \qquad F(t)=k_pP(t).

This is a deliberately compact model. It assumes well-mixed compartments, nonnegative counts, no direct input into the mature compartment, and no mature-cell proliferation. The constant-rate version further assumes that maturation and mature-cell loss rates do not vary with time or cell count. If there are additional pathways, the balance must include their fluxes rather than forcing them into these two terms.

What the Mature-Cell Slope Can and Cannot Tell You

If $k_m$ is constant, the balance identifies the instantaneous direction of change. When $F>k_mM$, the mature-cell count is increasing; when $F<k_mM$, it is decreasing. Equality means only that the instantaneous slope is zero. It is a necessary condition for a differentiable local peak, but it is not sufficient: the slope must also change from positive to negative. A change from negative to positive instead indicates a local minimum.

At a steady state, the balance gives $F=k_mM$. But during a transient, that equality may occur briefly while the system is moving through a peak or a trough. Therefore, a single observation of $F$ and $M$ cannot establish that a peak has occurred. Trajectories before and after the zero-slope time, or a justified model of the trajectory, are needed.

Diagnostic identity: If $M(t)>0$, the apparent mature-cell loss rate implied by a measured maturation flux and mature-cell slope is $$ k_{\mathrm{app}}(t)=\frac{F(t)-dM(t)/dt}{M(t)}. $$ Under the constant first-order loss model, $k_{\mathrm{app}}(t)=k_m$. A time-varying estimate may indicate changing loss, measurement error, omitted pathways, or a poor estimate of the derivative; it does not by itself identify which explanation is correct.

The equation also explains why mature-cell counts alone can be ambiguous. The same observed $dM/dt$ can result from a larger influx paired with larger loss, or a smaller influx paired with smaller loss. Independent information about the maturation flux, or a strong and testable model for that flux, is needed to separate the mechanisms.

Worked Examples

Worked Example: Checking a Candidate Peak With Fluxes

Suppose a hypothetical mature-cell population has $M=400$ cells and a measured maturation influx of $F=40$ cells per day. Assume first-order loss with $k_m=0.100$ per day. Determine the instantaneous slope and what additional evidence would establish a local peak.

The mature-cell loss flux is $$ L=k_mM=(0.100\ \text{day}^{-1})(400\ \text{cells})=40\ \text{cells/day}. $$ Therefore, $$ \frac{dM}{dt}=F-L=40-40=0\ \text{cells/day}. $$ This is a zero-slope time. It is consistent with a peak, but does not prove one. For example, if the slope was $+5$ cells per day just before this time and $-4$ cells per day just after it, the trajectory changes from rising to falling and has a local peak. If the signs were $-5$ and $+4$ cells per day, it would instead have a local minimum. Measurements or model predictions on both sides of the zero-slope time are needed to distinguish these cases.

Worked Example: Mature-Cell Overshoot After a Finite Input Pulse

Consider a hypothetical population initially at steady state with $P=100$ cells, $k_p=1.00$ per day, $M=200$ cells, and $k_m=0.500$ per day. The baseline input is $u=100$ cells per day. Input rises to $200$ cells per day for 2 days, then returns to baseline. Calculate the excess mature-cell count at the end of the pulse and the time and height of its later peak.

Write $p=P-100$ and $m=M-200$ for the excess counts above baseline. During the pulse, the excess precursor input is $100$ cells per day, so $dp/dt=100-p$ with $p(0)=0$. Its solution is $$ p(t)=100(1-e^{-t}). $$ Because the excess maturation influx is $p$ cells per day, $dm/dt=p-0.500m$, with $m(0)=0$. Solving this balance during the pulse gives $$ m(t)=200(1-e^{-0.500t})^2. $$ At day 2, $$ p(2)=100(1-e^{-2})=86.47\ \text{cells}, $$ and $$ m(2)=200(1-e^{-1})^2=79.92\ \text{cells}. $$ The mature count is therefore $M(2)=279.92$ cells, above its baseline of 200.

After the pulse, excess precursor input is zero, so $p(\tau)=p(2)e^{-\tau}$, where $\tau$ is time in days since the pulse ended. The mature-cell balance remains $dm/d\tau=p-0.500m$. Its solution, using the day-2 value as the initial condition, is $$ m(\tau)=\bigl(m(2)+2p(2)\bigr)e^{-0.500\tau}-2p(2)e^{-\tau} =252.84822e^{-0.500\tau}-172.93294e^{-\tau}. $$ The coefficients retain more digits than the reported counts to avoid rounding inconsistencies. At the post-pulse peak, $dm/d\tau=0$, so $p=0.500m$. Substitution gives $$ p(2)e^{-\tau} =0.500\left[\bigl(m(2)+2p(2)\bigr)e^{-0.500\tau}-2p(2)e^{-\tau}\right], $$ and hence $$ \tau=-2\ln\left(\frac{m(2)+2p(2)}{4p(2)}\right)=0.6265\ \text{days}. $$ At that time, $m=2p\approx92.42$ cells, so the total mature-cell count is $M\approx292.42$ cells. The peak is above baseline even though precursor excess is already declining: the precursor compartment continues to supply mature cells while mature-cell loss removes them more slowly than that influx at the peak. The slope is positive just after the pulse ends and changes to negative at this calculated maximum, confirming a peak rather than relying on the zero-slope condition alone.

Worked Example: Distinguishing a Loss-Rate Change From Delayed Influx

Suppose a hypothetical population is initially at equilibrium with $M=400$ cells and $F=40$ cells per day. The original loss rate is $0.100$ per day. At time zero, an independent flux measurement still gives $F=40$ cells per day, but a reliable local estimate of the mature-cell slope is $+20$ cells per day. Is this consistent with unchanged loss?

The balance implies a loss flux of $$ L=F-\frac{dM}{dt}=40-20=20\ \text{cells/day}. $$ The apparent loss rate is therefore $$ k_{\mathrm{app}}=\frac{L}{M}=\frac{20\ \text{cells/day}}{400\ \text{cells}}=0.0500\ \text{day}^{-1}. $$ That is half the original rate. If loss had remained at $0.100$ per day, its flux would still be $40$ cells per day and the predicted slope with $F=40$ would be zero, not $+20$. Given the stated measurements and model assumptions, this pattern supports a reduction in mature-cell loss, not a delayed increase in maturation influx. It does not prove that the biological loss rate changed: unmeasured direct input or another omitted pathway could also invalidate the balance used.

Diagnostics, Alternatives, and Limitations

A useful diagnosis compares the time courses of precursor count, maturation influx, mature-cell count, and mature-cell slope. Under the simple model, an input increase first changes $P$, which changes $F=k_pP$; the mature-cell response follows through its balance. During an overshoot caused by delayed maturation, the measured flux may remain above mature-cell loss after precursor input has started to fall. The mature-cell count can then rise until the flux and loss become equal, and it falls once flux becomes smaller than loss. In contrast, if $F$ is unchanged but $M$ begins rising faster than the unchanged-loss model predicts, the balance points to a change in loss or an omitted flux.

For constant $k_m$, the mature-cell trajectory can be written as a weighted accumulation of past influx:

M(t)=M(0)e^{-k_mt}+\int_0^t F(s)e^{-k_m(t-s)}\,ds.

This follows by multiplying $dM/dt+k_mM=F(t)$ by the integrating factor $e^{k_mt}$, integrating from 0 to $t$, and dividing by $e^{k_mt}$. Each past influx contributes to the current count, reduced by the fraction expected to remain after subsequent loss. This makes the lag explicit: mature-cell counts reflect both current flux and the history of earlier flux.

In practice, diagnostics can fail when measurements are too sparse to locate the slope change, when count noise makes finite-difference slopes unreliable, or when flux estimates are derived from the same data under the assumptions being tested. A value of $k_{\mathrm{app}}$ that varies with time is a warning to examine uncertainty, rate variation, unmeasured input, and model structure; it is not an automatic demonstration of time-varying loss. Estimating derivatives from noisy counts can amplify measurement error, so a fitted trajectory or repeated measurements may be more stable than a raw difference between adjacent counts.

Directly measuring maturation flux is often more informative than fitting mature counts alone, but it may require labeling or other specialized measurements. A model with an explicit transit delay or several maturation stages may better represent a broad maturation-time distribution than one precursor compartment. Such models also require more parameters and more informative data. A simpler alternative is to fit mature-cell counts with an effective input and loss model, but without independent information on input or flux, the two mechanisms can be difficult or impossible to distinguish. Model comparisons should therefore be guided by available measurements, biological plausibility, and whether the added detail can be identified.

The first-order model is limited when loss depends on cell density, varies over time, or differs across mature-cell subpopulations. It also does not represent mature-cell proliferation or direct mature-cell input. In those settings, use an explicit balance that includes the relevant pathways, and avoid interpreting $k_{\mathrm{app}}$ as a biological loss rate unless its assumptions are supported.

Common Mistake

Pitfall: “The mature-cell slope is zero when $F=k_mM$, so the count must be at a peak.” The equality establishes zero instantaneous slope, not a peak. Check that the slope is positive before the time and negative after it; the opposite sign change indicates a minimum. Also verify that the assumed loss law and measured flux account for the relevant pathways before attributing the turning point to maturation delay.

Check Your Understanding

Answer each question, then compare with the solutions below.

Under first-order mature-cell loss, $F=30$ cells per day, $M=300$ cells, and $k_m=0.0800$ per day. What is $dM/dt$?
At a time when $F=k_mM$, what additional evidence is needed to call the time a local peak?
If measured $F=50$ cells per day, $M=500$ cells, and $dM/dt=10$ cells per day, calculate $k_{\mathrm{app}}$.
Why can a mature-cell count trajectory by itself fail to distinguish a change in maturation influx from a change in mature-cell loss?
In the finite-pulse example, why can $M$ continue rising after precursor excess starts to decline?

Solutions. (1) Loss is $k_mM=(0.0800)(300)=24.0$ cells per day, so $dM/dt=30-24.0=6.00$ cells per day. (2) Verify that the slope changes from positive to negative around the zero-slope time; zero slope alone could also occur at a minimum. (3) The implied loss is $F-dM/dt=50-10=40$ cells per day, so $k_{\mathrm{app}}=40/500=0.0800$ per day. (4) The balance contains the difference $F-L$; different combinations of influx and loss can yield the same slope and count. Independent flux information or additional justified assumptions are needed. (5) Declining precursor excess can still produce a maturation flux greater than mature-cell loss. The mature count continues to rise while $F>k_mM$, and reaches its peak only when the slope changes from positive to negative through the zero-slope condition.