The Two-Compartment Kinetic Model

Why Represent RNA in Two Compartments?

RNA production and processing involve distinct stages. Newly transcribed RNA is unspliced; splicing transfers RNA into the spliced compartment, where it may later be degraded. A model that keeps these stages separate can represent the fact that transcription changes the unspliced amount first, while splicing both removes unspliced RNA and supplies spliced RNA.

The two-compartment kinetic model describes these amounts as functions of time. It is a mechanistic accounting model: each change is attributed to a specified flux. The equations do not, by themselves, establish that a cell population is following a particular biological trajectory. Their interpretation depends on the measurement scale, the time unit, and the assumptions used to define the fluxes.

Definition: Let $u(t)$ denote the amount of unspliced RNA and $s(t)$ the amount of spliced RNA for a gene at time $t$. In the basic two-compartment model, $\alpha$ is the transcription rate, $\beta$ is the splicing rate constant, and $\gamma$ is the degradation rate constant for spliced RNA. The model is $$ \frac{du}{dt}=\alpha-\beta u,\qquad \frac{ds}{dt}=\beta u-\gamma s. $$ Here $\alpha$ has units of RNA amount per time, while $\beta$ and $\gamma$ have units of inverse time. Thus both derivatives have units of RNA amount per time.

The amounts may be molecule counts in an idealized cell or quantities on a specified abundance scale. If the data have been normalized, the amount units must be understood as units on that scale; the rate units change accordingly. Whatever scale is used, $u$, $s$, and the rates must be expressed consistently.

Assumptions Behind the Equations

The equations describe a particular, deliberately simple flux structure. In the basic model, transcription adds unspliced RNA at a constant rate $\alpha$ over the time interval of interest. Splicing removes unspliced RNA at a rate proportional to its current amount, with flux $\beta u$. That same flux enters the spliced compartment. Spliced RNA is removed by degradation at a rate proportional to its amount, with flux $\gamma s$.

This specification makes several assumptions explicit:

Well-mixed compartments: a single amount $u(t)$ and $s(t)$ adequately summarize the RNA in each compartment.
First-order splicing and degradation: the corresponding flux is proportional to the compartment amount, with constant proportionality rate over the interval.
Transfer by splicing: splicing moves RNA from the unspliced compartment to the spliced one; it does not remove RNA from the combined two-compartment total.
No independent unspliced-RNA loss in the basic model: the only modeled exit from the unspliced compartment is splicing.
Continuous-time description: changes are represented by rates, so the differential equations approximate the evolution of the modeled amounts rather than individual discrete molecular events.

The constant-transcription assumption is a modeling choice, not a claim that transcription is biologically unchanging. If transcription varies over time, it can be written as $\alpha(t)$. The same flux accounting then gives $du/dt=\alpha(t)-\beta u$, while the spliced equation remains $ds/dt=\beta u-\gamma s$ if the other assumptions still hold.

Deriving the Coupled ODEs

The equations follow from changes over a short interval $\Delta t$. During this interval, transcription contributes approximately $\alpha\Delta t$ units of unspliced RNA. Splicing removes approximately $\beta u(t)\Delta t$ from the unspliced amount and adds that same amount to the spliced amount. Degradation removes approximately $\gamma s(t)\Delta t$ from the spliced amount. These are expected or smooth-model fluxes; they need not be integer counts over a short interval.

\begin{aligned} u(t+\Delta t)-u(t) &=\alpha\Delta t-\beta u(t)\Delta t+o(\Delta t),\\ s(t+\Delta t)-s(t) &=\beta u(t)\Delta t-\gamma s(t)\Delta t+o(\Delta t). \end{aligned}

The remainder $o(\Delta t)$ denotes terms that become negligible relative to $\Delta t$ as $\Delta t$ approaches zero. Divide each balance by $\Delta t$, then take the limit as $\Delta t$ tends to zero. By the definition of a time derivative, the resulting equations are

\frac{du}{dt}=\alpha-\beta u,\qquad \frac{ds}{dt}=\beta u-\gamma s.

This derivation also clarifies why the equations are coupled. The spliced-RNA derivative depends on $u$, because unspliced RNA is the source of the splicing flux. The unspliced derivative does not depend on $s$ in this basic model: there is no modeled reverse transfer from spliced to unspliced RNA.

Key interpretation: The term $\alpha$ adds unspliced RNA; $-\beta u$ removes it through splicing; $+\beta u$ supplies spliced RNA through that same splicing flux; and $-\gamma s$ removes spliced RNA by degradation. The paired splicing terms have equal magnitude because they represent one transfer between compartments.

Worked Example: Calculating Both Instantaneous Changes

Suppose a hypothetical gene has $u=24$ RNA units and $s=50$ RNA units at a particular time. Let $\alpha=10$ RNA units per hour, $\beta=0.25$ per hour, and $\gamma=0.10$ per hour. Assume the basic model's flux structure and constant rates apply at that time.

For unspliced RNA, the transcription input is $10$ units per hour and the splicing output is $\beta u=(0.25\ \text{hour}^{-1})(24\ \text{units})=6$ units per hour. Therefore,

\frac{du}{dt}=\alpha-\beta u =10-(0.25)(24) =10-6 =4\ \text{RNA units per hour}.

For spliced RNA, splicing supplies $6$ units per hour and degradation removes $\gamma s=(0.10\ \text{hour}^{-1})(50\ \text{units})=5$ units per hour. Thus,

\frac{ds}{dt}=\beta u-\gamma s =(0.25)(24)-(0.10)(50) =6-5 =1\ \text{RNA unit per hour}.

At this instant, the modeled unspliced and spliced amounts are both increasing. These are instantaneous rates of change, not predictions that either amount will increase at the same rate over a long interval.

What the Splicing Terms Tell Us

Splicing is an internal transfer, so it appears with opposite signs in the two equations. Adding the equations cancels the transfer terms:

\frac{d(u+s)}{dt} =\frac{du}{dt}+\frac{ds}{dt} =(\alpha-\beta u)+(\beta u-\gamma s) =\alpha-\gamma s.

This identity is a useful accounting check. In the basic model, transcription increases the total $u+s$, and degradation of spliced RNA decreases it. Splicing alone does not change the total. If a proposed equation adds splicing to both compartments, or subtracts it from both, it does not represent the transfer specified by this model.

The derivative signs also have a precise local interpretation. If $\alpha-\beta u$ is positive, unspliced RNA is increasing at that instant; if it is negative, unspliced RNA is decreasing. Similarly, the sign of $\beta u-\gamma s$ describes the instantaneous change in spliced RNA. A zero derivative means the modeled amount has zero instantaneous slope at that time; it does not mean that the amount is zero or that it must remain unchanged afterwards.

Worked Example: One Compartment Can Increase While the Other Is Unchanged

Consider a hypothetical cell with $u=8$ RNA units and $s=20$ RNA units. Suppose $\alpha=6$ units per hour, $\beta=0.50$ per hour, and $\gamma=0.20$ per hour. Assume the basic model applies.

The splicing flux is $(0.50\ \text{hour}^{-1})(8\ \text{units})=4$ units per hour. The transcription input is $6$ units per hour, so

\frac{du}{dt}=6-4=2\ \text{RNA units per hour}.

The degradation flux is $(0.20\ \text{hour}^{-1})(20\ \text{units})=4$ units per hour. Splicing supplies the same $4$ units per hour, giving

\frac{ds}{dt}=4-4=0\ \text{RNA units per hour}.

Thus the spliced amount has zero instantaneous rate of change even as the unspliced amount increases at $2$ units per hour. The zero results from equal input and removal fluxes at this instant; it does not imply that transcription or splicing has stopped.

Extensions and Model Boundaries

The basic model excludes independent degradation of unspliced RNA. If the biology or measurement process calls for that loss, introduce a separate unspliced degradation rate constant $\delta$, measured in inverse time. Its flux is $\delta u$, so the extended equations are

\frac{du}{dt}=\alpha-(\beta+\delta)u,\qquad \frac{ds}{dt}=\beta u-\gamma s.

Only the unspliced equation changes: degradation removes unspliced RNA without supplying spliced RNA. The total-amount balance becomes $d(u+s)/dt=\alpha-\delta u-\gamma s$. This extension should not be added automatically. It introduces another rate and should be used only when it is part of the chosen biological model and can be supported by the available information.

Worked Example: Accounting for Unspliced-RNA Degradation

Suppose a hypothetical gene has $u=20$ units, $s=12$ units, $\alpha=9$ units per hour, $\beta=0.30$ per hour, $\gamma=0.15$ per hour, and an independently specified unspliced degradation rate $\delta=0.05$ per hour. Use the extended model.

Unspliced RNA leaves through both splicing and degradation. The splicing flux is $(0.30)(20)=6$ units per hour, while unspliced degradation is $(0.05)(20)=1$ unit per hour. Consequently,

\frac{du}{dt} =9-(0.30+0.05)(20) =9-7 =2\ \text{RNA units per hour}.

Spliced RNA still receives the $6$-unit-per-hour splicing flux. Its degradation flux is $(0.15)(12)=1.8$ units per hour, so

\frac{ds}{dt} =(0.30)(20)-(0.15)(12) =6-1.8 =4.2\ \text{RNA units per hour}.

The total derivative is $2+4.2=6.2$ units per hour. The extended total-balance formula gives the same result: $9-(0.05)(20)-(0.15)(12)=9-1-1.8=6.2$. This agreement checks that the unspliced degradation term removes RNA rather than transferring it.

Diagnostics, Alternatives, and Limitations

Before interpreting a fitted or specified model, check that each term matches its biological process and that all units are compatible. The products $\beta u$, $\gamma s$, and, when included, $\delta u$ must have units of RNA amount per time, just like $\alpha$. Nonnegative amounts and nonnegative rate constants are natural for this model. A negative fitted rate or a sign pattern that treats degradation as an input signals a problem with parameter constraints, coding, or interpretation.

Other warning signs include unexplained changes in rates across the interval, strong departures from first-order flux assumptions, or evidence that transcription is changing when the model assumes constant $\alpha$. A mismatch can also arise when observed counts are noisy measurements of underlying amounts. The differential equations describe smooth changes in modeled RNA abundance; raw counts need not follow these equations exactly at every observation.

The basic model is more informative than a one-compartment model when the distinction between unspliced and spliced RNA matters, because it represents splicing as a transfer and degradation as a separate loss. A model with explicit unspliced degradation is a plausible alternative when precursor loss is required, but it has an additional rate parameter. A discrete-time accounting model can instead track changes over finite sampling intervals; it must specify how fluxes operate during each interval and should not be mistaken for the continuous-time equations without an approximation. These approaches differ in assumptions and detail, not merely notation.

The model's simplicity is also a limitation. Constant rates and two well-mixed compartments may not capture transcriptional changes, multiple processing stages, or cell-to-cell biological variability. In single-cell data, measured counts are snapshots and are subject to sampling and technical variation; they are not direct observations of a cell's entire time course. The equations define a kinetic description, but deciding whether its assumptions are adequate requires biological context and appropriate data.

Common Mistakes

Counting splicing as loss from the whole system. Splicing removes RNA from $u$, but the same flux enters $s$. It cancels when the two equations are added.
Putting transcription directly into both equations. In this model, transcription enters the unspliced compartment. Spliced RNA is supplied through the $\beta u$ splicing flux.
Confusing a rate constant with a flux. $\beta$ has units of inverse time; the splicing flux is $\beta u$, with units of RNA amount per time.
Assuming a zero derivative means no molecular activity. A zero derivative can result from equal incoming and outgoing fluxes, as in the second worked example.
Adding unspliced degradation without changing the model statement. The basic model omits that loss. If it is included, specify $\delta$ and subtract $\delta u$ from the unspliced balance.

Key takeaway: The basic two-compartment model is a flux balance: transcription adds to $u$, splicing transfers $\beta u$ from $u$ to $s$, and degradation removes $\gamma s$ from $s$. The signs, units, and shared splicing flux should all agree with that accounting.

Check Your Understanding

Answer each question, then compare your reasoning with the concise solutions below.

For the basic model, what process is represented by each of $\alpha$, $-\beta u$, $+\beta u$, and $-\gamma s$?
If $u=10$ units, $s=15$ units, $\alpha=5$ units per hour, $\beta=0.20$ per hour, and $\gamma=0.10$ per hour, calculate both instantaneous derivatives.
Why does splicing cancel from the derivative of $u+s$ in the basic model?
What additional term belongs in the unspliced equation if unspliced RNA degrades independently at rate constant $\delta$?
A researcher says that $ds/dt=0$ proves no RNA is being spliced. Explain the error.

Solutions. (1) $\alpha$ is transcription input to unspliced RNA; $-\beta u$ is its removal by splicing; $+\beta u$ is the corresponding supply to spliced RNA; and $-\gamma s$ is spliced-RNA degradation. (2) $du/dt=5-(0.20)(10)=3$ units per hour, and $ds/dt=(0.20)(10)-(0.10)(15)=0.5$ units per hour. (3) Splicing removes and supplies the same flux, so $-\beta u+\beta u=0$ in the sum. (4) Subtract $\delta u$, giving $du/dt=\alpha-\beta u-\delta u$. (5) A zero spliced derivative can mean the splicing input equals degradation output; here it would not imply either flux is zero.