Why Conserved Totals Matter
In a closed reaction system, the amounts of individual species can change while certain totals remain fixed. Those conservation laws are more than bookkeeping: they reveal which state variables are independent, whether multiple state measurements provide distinct information, and which parameter conclusions are supported by the observations. The key distinction is between redundancy in the state representation and ambiguity about parameters. The first can occur even when the reaction parameters are identifiable; the second means different parameter values produce the same ideal observations.
Consider a reversible binding reaction in a fixed volume, with free species \(A\) and \(B\) forming complex \(C\). Assume a closed system: there is no synthesis, degradation, transport, or external input, and the reaction is the only process changing these species. Under mass-action kinetics, the forward and reverse reaction rates are \(k_{\mathrm{on}}AB\) and \(k_{\mathrm{off}}C\), respectively.
Deriving the Conservation Laws
Write the state vector as \(x=(A,B,C)^\mathsf{T}\). The reaction changes the states in the proportions \((-1,-1,+1)\): one molecule of each free species is consumed and one complex is formed. The model is therefore
A linear combination \(l^\mathsf{T}x\) is conserved if its derivative is zero for every possible reaction rate. Since \(\dot{x}=s\,v\), where \(s=(-1,-1,1)^\mathsf{T}\) is the stoichiometric vector and \(v=k_{\mathrm{on}}AB-k_{\mathrm{off}}C\), this requires \(l^\mathsf{T}s=0\). Two independent choices are \(l_1=(1,0,1)^\mathsf{T}\) and \(l_2=(0,1,1)^\mathsf{T}\), giving the totals
where \(A_T\) and \(B_T\) are fixed by the initial amounts. Thus \(A=A_T-C\) and \(B=B_T-C\). The three named species are not three independent dynamic degrees of freedom: once the two totals are specified, one state determines the other two.
Substitution into the complex equation gives a one-state model:
For physically feasible states, \(0\leq C\leq\min(A_T,B_T)\). This reduced form is equivalent to the full three-state model when the initial states satisfy the conservation laws. It is not an approximation: it is a coordinate reduction to the independent state.
Worked Example: Recovering Kinetic Rates With Known Totals
Suppose a hypothetical, noiseless continuous record of complex concentration has \(A_T=10\ \mu\mathrm{M}\), \(B_T=6\ \mu\mathrm{M}\), and \(C(0)=0\ \mu\mathrm{M}\). Assume the closed mass-action model above, with positive constant \(k_{\mathrm{on}}\) and \(k_{\mathrm{off}}\). The initial derivatives inferred from the ideal record are \(C'(0)=0.12\ \mu\mathrm{M}/\mathrm{min}\) and \(C''(0)=-0.006\ \mu\mathrm{M}/\mathrm{min}^2\). Show how they determine both rates.
At \(C=0\), the free concentrations are \(A=10\ \mu\mathrm{M}\) and \(B=6\ \mu\mathrm{M}\). The initial slope is therefore
To use the curvature, write the reduced right-hand side as \(f(C)\). Differentiating \(\dot C=f(C)\) with respect to time gives \(C''=f'(C)C'\), where
At \(C(0)=0\), the supplied derivatives imply
Using \(k_{\mathrm{on}}=0.002\ \mu\mathrm{M}^{-1}\mathrm{min}^{-1}\), \(16k_{\mathrm{on}}=0.032\ \mathrm{min}^{-1}\), so \(k_{\mathrm{off}}=0.05-0.032=0.018\ \mathrm{min}^{-1}\). Under the stated model and known totals, the initial slope and curvature recover a unique positive rate pair. This is an ideal structural argument; estimating derivatives from finite, noisy measurements is a different and generally harder problem.
Redundant Measurements Are Not Automatically Redundant Parameters
If both totals are known, measuring \(C\) determines \(A\) and \(B\) through \(A=A_T-C\) and \(B=B_T-C\). Conversely, measuring \(A\) and \(C\) determines \(B\). In an exact model, reporting all three state trajectories does not create three independent output trajectories. It does, however, provide useful checks: departures from the mass-balance equations can signal measurement error or model mismatch.
Worked Example: Identifying an Algebraically Redundant State
At one time in a hypothetical closed binding experiment, the known totals are \(A_T=10\ \mu\mathrm{M}\) and \(B_T=6\ \mu\mathrm{M}\). Measurements report \(C=2.0\ \mu\mathrm{M}\), \(A=8.0\ \mu\mathrm{M}\), and \(B=4.0\ \mu\mathrm{M}\). Check the mass balances and determine which measurements are redundant in the exact model.
The two conserved totals calculated from the reported states are \(A+C=8.0+2.0=10.0\ \mu\mathrm{M}\) and \(B+C=4.0+2.0=6.0\ \mu\mathrm{M}\). Both match their specified totals. Given \(C=2.0\ \mu\mathrm{M}\) alone, the model predicts \(A=10.0-2.0=8.0\ \mu\mathrm{M}\) and \(B=6.0-2.0=4.0\ \mu\mathrm{M}\). Thus the three exact outputs are consistent, but only one is an independent state trajectory once the totals are known.
This algebraic dependence does not show that \(k_{\mathrm{on}}\) and \(k_{\mathrm{off}}\) are indistinguishable. Their identifiability depends on how the independent state evolves and what information the observations contain. For example, the time-course derivatives in the preceding example distinguish the rates even though the three state outputs are constrained by two conservation equations.
Distinguishing State Reduction From Parameter Ambiguity
A genuine parameter ambiguity occurs when distinct parameter values yield the same ideal observations under the specified experimental conditions. Conservation alone does not establish such an ambiguity. It reduces the number of independent state coordinates; whether parameters can be distinguished still depends on the observation type and the dynamics. Equilibrium-only measurements illustrate the difference.
Worked Example: Equilibrium Data Identify a Rate Ratio, Not Both Rates
Suppose a hypothetical experiment measures only the equilibrium state, with known totals \(A_T=10\ \mu\mathrm{M}\), \(B_T=6\ \mu\mathrm{M}\), and \(C^*=2.0\ \mu\mathrm{M}\). Assume the system has reached equilibrium under the reversible mass-action model. Find the parameter combination determined by this observation and show why the individual rates are not determined.
The equilibrium free concentrations are \(A^*=10-2=8\ \mu\mathrm{M}\) and \(B^*=6-2=4\ \mu\mathrm{M}\). At equilibrium, forward and reverse rates are equal:
The equilibrium observation identifies \(K_D=k_{\mathrm{off}}/k_{\mathrm{on}}=16\ \mu\mathrm{M}\), but not either rate separately. For example, \(k_{\mathrm{on}}=0.002\ \mu\mathrm{M}^{-1}\mathrm{min}^{-1}\), \(k_{\mathrm{off}}=0.032\ \mathrm{min}^{-1}\) and \(k_{\mathrm{on}}=0.005\ \mu\mathrm{M}^{-1}\mathrm{min}^{-1}\), \(k_{\mathrm{off}}=0.080\ \mathrm{min}^{-1}\) have the same ratio and satisfy the same equilibrium condition. The rate pairs differ, so this is actual parameter ambiguity for equilibrium-only data, not a consequence of having written three states instead of one.
Practical Checks and Model Choices
The full three-state equations and the reduced one-state equation are alternative representations of the same closed binding model. The full form can be convenient when state-specific equations or measurements are central. The reduced form makes the independent dynamics explicit and can simplify analysis. Whichever form is used, initial conditions must satisfy \(A(0)+C(0)=A_T\) and \(B(0)+C(0)=B_T\); otherwise the initial state is outside the specified closed-system model.
In ideal data, conservation residuals \(A(t)+C(t)-A_T\) and \(B(t)+C(t)-B_T\) should be zero at every time. With measured data they need not be exactly zero. Large, systematic departures may indicate an omitted process, an incorrect total, or a measurement problem. Small departures can arise from measurement error. A discrepancy is a diagnostic, not direct proof of which explanation is correct.
When measurement errors are present, algebraically related observations should not be treated as independent exact equations. Their errors and any shared measurements affect how much information they provide. Multiple noisy measurements can still be useful, but their value depends on the observation-error model; the exact conservation identity does not imply that repeated noisy readings are useless.
A common mistake is to see three named species and infer that three independent state trajectories automatically supply three independent constraints on the parameters. The correction is to derive the conserved totals first, determine how many states are independent, and then assess the information in the remaining dynamics and observation conditions. A second mistake is to treat an equilibrium rate-ratio ambiguity as proof that the state representation is redundant. The equilibrium example shows a parameter limitation; the known-total time-course example shows that a reduced state representation can still support identification of distinct rates.
Use the stoichiometric vector to specify how each reaction changes each state.
Find linear combinations whose products with the stoichiometric vector are zero, then evaluate their constants from the initial state.
Substitute the totals to identify independent states. Separately ask whether the available ideal observations distinguish the parameters.
Check Your Understanding
Answer each question, then compare with the brief solutions.
- For the reaction \(A+B\rightleftharpoons C\), state the two conserved totals.
- If \(A_T=12\ \mu\mathrm{M}\), \(B_T=5\ \mu\mathrm{M}\), and \(C=1.5\ \mu\mathrm{M}\), calculate \(A\) and \(B\).
- With both totals known, does observing all three exact states provide three independent state trajectories? Explain.
- For equilibrium-only data, what parameter combination follows from \(k_{\mathrm{on}}A^*B^*=k_{\mathrm{off}}C^*\)?
- What does a nonzero measured conservation residual establish?
Solutions. (1) \(A+C=A_T\) and \(B+C=B_T\). (2) \(A=12-1.5=10.5\ \mu\mathrm{M}\) and \(B=5-1.5=3.5\ \mu\mathrm{M}\). (3) No. Any one state determines the other two when both totals are known, so the exact outputs are algebraically constrained. (4) It gives \(k_{\mathrm{off}}/k_{\mathrm{on}}=A^*B^*/C^*\), not the two rates separately. (5) It establishes a discrepancy from exact mass balance in the measured values; by itself it does not identify whether measurement error, an incorrect total, or omitted model processes caused that discrepancy.