Elemental Balance Checks for Biochemical Reaction Networks

Why Check Elemental Balance?

A reaction network can produce plausible concentration profiles and still encode an impossible chemical transformation. If a reaction creates nitrogen atoms, loses carbon atoms, or changes the number of phosphorus atoms without an external source, its stoichiometry is inconsistent with the molecular formulas assigned to its species. An elemental audit is a direct way to find such errors before interpreting model behavior.

The check is local to each reaction: count each element on the reactant side and on the product side, including stoichiometric coefficients. This is different from asking whether the network has a conserved pool of a particular species or molecular group. A reaction may conserve every element while converting one metabolite into another, and a closed network may also have conserved totals that combine multiple species.

Definition: Let $q$ be the number of modeled species and $r$ the number of reactions. The element-composition matrix $A$ has one row for each element being checked and one column for each species. Entry $A_{ki}$ is the number of atoms of element $k$ in one molecule of species $i$. The stoichiometric matrix $S$ has one row per species and one column per reaction, with reactant coefficients negative and product coefficients positive. A reaction network is elementally balanced when $$ A S = 0. $$

This condition assumes that each species has a specified molecular formula and that the reaction list accounts for all species participating in the transformations being checked. It tests atom conservation for the elements represented in $A$, not chemical plausibility in every other respect. Charge balance, for example, is a separate check. Formula conventions must also be consistent: if species are represented in particular protonation states, those formulas must be used consistently, and any water or proton participating in a reaction must be included when relevant.

From Molecular Formulas to a Balance Test

For a single reaction, write the stoichiometric coefficients in a vector $s_j$, the $j$-th column of $S$. Multiplying $A s_j$ adds the product-side atom counts and subtracts the reactant-side atom counts, element by element. Thus, the $k$-th entry of $A s_j$ is the product-minus-reactant difference in atoms of element $k$, per reaction event. Balance requires every entry to equal zero.

(A s_j)_k = \sum_{i=1}^{q} A_{ki} S_{ij} = \text{product atoms of element } k - \text{reactant atoms of element } k

The matrix test carries out this count for every element and every reaction at once. A nonzero entry identifies both a reaction (the corresponding column) and an element (the corresponding row) that fails the check. The sign indicates whether the listed products contain more or fewer atoms than the listed reactants.

Worked Example: Glucose Conversion to Lactate

Suppose a hypothetical closed reaction is written as glucose $\rightarrow$ 2 lactate, using the neutral molecular formulas glucose $C_6H_{12}O_6$ and lactic acid $C_3H_6O_3$. Check carbon, hydrogen, and oxygen.

The reactant counts are $C=6$, $H=12$, and $O=6$. One lactic acid molecule has $C=3$, $H=6$, and $O=3$, so the two product molecules contain $C=2(3)=6$, $H=2(6)=12$, and $O=2(3)=6$.

With species ordered as glucose, lactate, the stoichiometric column is $s=(-1,2)^\mathsf{T}$. For element rows ordered $C,H,O$, the composition matrix is

A = \begin{pmatrix} 6 & 3 \\ 12 & 6 \\ 6 & 3 \end{pmatrix}, \qquad A s = \begin{pmatrix} -6+2(3) \\ -12+2(6) \\ -6+2(3) \end{pmatrix} = \begin{pmatrix} 0\\0\\0 \end{pmatrix}.

The written reaction conserves all three listed elements. This is an atom-count result for the stated formulas and stoichiometry; it does not by itself establish that the reaction occurs by a particular mechanism or at a particular rate.

Worked Audits of Biochemical Reactions

Worked Example: Checking ATP Hydrolysis

For this hypothetical audit, use the neutral-form formulas ATP $C_{10}H_{16}N_5O_{13}P_3$, water $H_2O$, ADP $C_{10}H_{15}N_5O_{10}P_2$, and phosphoric acid $H_3PO_4$. Check the reaction ATP + water $\rightarrow$ ADP + phosphoric acid.

On the reactant side, ATP plus water gives $C=10$, $H=16+2=18$, $N=5$, $O=13+1=14$, and $P=3$. On the product side, ADP plus phosphoric acid gives $C=10$, $H=15+3=18$, $N=5$, $O=10+4=14$, and $P=2+1=3$.

Using element order $C,H,N,O,P$, the product-minus-reactant residual is

\begin{pmatrix} 10-10\\ 18-18\\ 5-5\\ 14-14\\ 3-3 \end{pmatrix} = \begin{pmatrix} 0\\0\\0\\0\\0 \end{pmatrix}.

Every listed element balances. In a model using ionic or other protonation-state formulas, the molecular formulas and any explicit proton coefficients should instead match that representation. Do not mix formula conventions within one audit.

Worked Example: Finding an Incorrect Ammonia Coefficient

Consider glutamine hydrolysis with formulas glutamine $C_5H_{10}N_2O_3$, water $H_2O$, glutamic acid $C_5H_9NO_4$, and ammonia $NH_3$. First test the proposed reaction glutamine + water $\rightarrow$ glutamic acid + 2 ammonia.

The reactant totals in the order $C,H,N,O$ are $(5,10+2,2,3+1)=(5,12,2,4)$. The proposed product totals are $(5,9+2(0),1+2,4+2(0))=(5,9,3,4)$ for the atoms in the listed formulas, except that hydrogen on the products is $9+2(3)=15$. Thus the full product vector is $(5,15,3,4)$. The product-minus-reactant residual is $(0,3,1,0)$: the products contain three extra hydrogen atoms and one extra nitrogen atom.

The coefficient can be corrected by testing one ammonia molecule instead. The product totals become $C=5$, $H=9+3=12$, $N=1+1=2$, and $O=4$, matching the reactant totals. The balanced reaction under these formulas is glutamine + water $\rightarrow$ glutamic acid + ammonia. The residual pinpointed the problem rather than merely flagging that something was wrong.

Why a Balanced Network Has Conserved Element Totals

Elemental balance also gives a network-level result. Let $x$ be the vector of species amounts, and let $v$ be the vector of reaction rates, with rates defined in the directions of the columns of $S$. For a closed system described by $\dot{x}=Sv$, the total amounts of the elements are the entries of $Ax$. If $AS=0$, then

\frac{d(Ax)}{dt} = A\frac{dx}{dt} = A S v = 0.

Therefore $Ax$ is constant over time: every total elemental amount remains equal to its initial value. This conclusion follows regardless of how the reaction rates depend on the state, provided the closed-system balance equation and elemental balance condition hold.

Theorem: If a closed reaction network obeys $\dot{x}=Sv$ and its species formulas give $AS=0$, then $Ax(t)=Ax(0)$ for all times for which the model solution exists. For an open system with an external amount-flow vector $u$, $\dot{x}=Sv+u$, the elemental totals instead obey $\frac{d(Ax)}{dt}=Au$; they need not be constant when external flows carry elements into or out of the modeled system.

This derivation uses amounts. If states are expressed as concentrations in compartments with changing volumes, the amount represented by each concentration must be accounted for before applying a total-element statement. The compartment-volume bookkeeping is distinct from the atom-count test itself.

Audit Procedure, Diagnostics, and Limits

A reliable audit is simple to organize, but it depends on getting the model representation right. Use a consistent species list, record the chosen formula for each species, and compare the coefficients used in the reaction model with those in the written reaction. For every reaction, calculate product-minus-reactant counts for every element included in the check.

Set the species formulas.
Record the molecular formula for each modeled species and state the convention for protonation states. Identify any species whose formula is unknown or deliberately approximate.

Build the composition matrix.
Choose element rows, such as carbon, hydrogen, nitrogen, oxygen, phosphorus, and sulfur, then enter the atom count for each species in the corresponding matrix column.

Apply the reaction coefficients.
Use negative stoichiometric entries for reactants and positive entries for products. Compute $A S$ and inspect each nonzero residual by reaction and element.

Trace each residual to the model representation.
Check coefficients, formulas, and potentially omitted participants such as water, protons, or an external species. Recheck the reaction after any change.

A nonzero residual is a diagnostic, not an instruction to add a species automatically. It can indicate a transcription error, a wrong coefficient, an incorrect molecular formula, or an incomplete reaction description. In a biochemical model, some species may stand for pools or lumped entities without a single known formula. Their elemental composition cannot be checked exactly unless it is specified or otherwise constrained.

Elemental balance is also not identical to charge balance. A reaction can conserve the atoms represented in $A$ yet fail a charge audit if the assigned species charges and any explicit ionic participants do not balance. Conversely, a charge-balanced equation can still create or lose atoms. These checks answer different questions and should be kept separate.

Check	What it tests	What it cannot establish alone
Elemental balance $AS=0$	Whether listed reactions conserve each included element	Whether rates, mechanisms, or the model structure are biologically correct
Conserved totals from stoichiometry	Whether particular combinations of species amounts remain constant in the network	Whether each reaction is atom-balanced under the assigned formulas
Charge balance	Whether listed electrical charge is conserved in a reaction representation	Whether carbon, nitrogen, or other atoms balance

The earlier material on conserved moieties uses stoichiometric relationships to identify totals that remain unchanged. Elemental bookkeeping supplies a chemically interpretable check on those relationships when molecular formulas are available. A formula-based audit is especially useful for locating missing or miscounted atoms, while a general stoichiometric conservation analysis can identify conserved combinations even when those combinations are not elemental totals.

Common Mistakes

Checking formulas but ignoring coefficients. Two product molecules contain twice the atoms of one molecule. Multiply every formula count by its stoichiometric coefficient.
Assuming a balanced reaction is biologically valid. $AS=0$ is necessary for atom conservation under the stated representation, but it does not validate the reaction rate law, direction, or mechanism.
Mixing protonation conventions. Formula differences among ionic and neutral forms can change hydrogen counts. Use a consistent representation and include explicit participants when required.
Treating an open system as closed. Inputs and outputs can change the amount of an element inside the modeled system. In that case the elemental change must be reconciled with the external flows.
Interpreting a nonzero residual as proof of one particular error. The residual localizes the imbalance, but the cause could be a coefficient, formula, omitted species, or the chosen system boundary.

Key takeaway: Molecular formulas turn an elemental audit into a matrix calculation: construct $A$, use the signed reaction stoichiometry $S$, and check whether $AS=0$. A zero residual confirms atom conservation for the elements and species represented; a nonzero residual identifies where the model representation needs investigation.

Check Your Understanding

Answer each question, then compare with the solutions below.

A hypothetical reaction is $2A\rightarrow B$, where $A$ has formula $C_2H_4O_2$ and $B$ has formula $C_4H_8O_4$. Is it balanced for carbon, hydrogen, and oxygen?
For species amounts $x$, reaction rates $v$, and $\dot{x}=Sv$, state the condition that makes all represented elemental totals constant in a closed system.
For $A+B\rightarrow C$, an audit gives a nitrogen residual of $-1$, where residual means product atoms minus reactant atoms. What does the sign mean?
Why can a reaction pass an elemental balance test but fail a charge balance test?

Solutions. (1) Yes. The reactant totals are $C=2(2)=4$, $H=2(4)=8$, and $O=2(2)=4$, which match $B$'s $C_4H_8O_4$. (2) The condition is $AS=0$. Then $\frac{d(Ax)}{dt}=ASv=0$, so $Ax$ is constant. (3) The products have one fewer nitrogen atom than the reactants for the written reaction; the nitrogen count is not conserved. (4) The elemental test counts atoms, not electrical charges. Species can contain the same numbers of each element while their assigned charges fail to sum to the same value on both sides.

Check	What it tests	What it cannot establish alone
Elemental balance \(AS=0\)	Whether listed reactions conserve each included element	Whether rates, mechanisms, or the model structure are biologically correct
Conserved totals from stoichiometry	Whether particular combinations of species amounts remain constant in the network	Whether each reaction is atom-balanced under the assigned formulas
Charge balance	Whether listed electrical charge is conserved in a reaction representation	Whether carbon, nitrogen, or other atoms balance