Diagnosing Why Published Measurements Disagree

Why Published Enzyme Activity Estimates Can Disagree

Two publications can report different enzyme activity values without either having made an arithmetic error. One may report a rate at a particular substrate concentration, while another reports an estimated maximum rate. They may also use different biological material, sample processing, normalization, or participant populations. Before treating the values as competing estimates of one quantity, identify what each measurement represents and how it was produced.

The task here is diagnosis: to determine which differences could plausibly account for the reported disagreement and what information is needed to assess them. A numerical difference by itself does not reveal its cause. Nor does matching the units guarantee that two estimates refer to the same endpoint, material, or population.

Definition: A measurement discrepancy is a difference between reported numerical results. A measurement disagreement is a discrepancy between results that are intended to represent the same defined quantity under sufficiently comparable conditions. A discrepancy can be real and reproducible even when the underlying estimates are not directly comparable.

For an enzyme activity estimate, first specify the enzyme, the biological material, the assay conditions, the calculated endpoint, and the normalization denominator. For example, “product formed per minute per milligram of microsomal protein at a stated substrate concentration” is more specific than “enzyme activity.” In a PBPK parameter record, this distinction matters because a number is only useful when its meaning and applicability are clear.

Separate the Sources of Difference

A practical diagnosis asks about five related sources. They are not mutually exclusive: two studies can differ in more than one at once.

Source	Questions to check	Why it matters
Assay design	Were the probe substrate, substrate concentration, cofactors, incubation time, and analytical method comparable?	The measured rate can depend on assay conditions, even when the same enzyme and material are used.
Biological material	Was the assay performed in recombinant enzyme, a tissue fraction, cells, or intact tissue?	These materials differ in enzyme abundance, other enzymes, binding components, and biological context.
Sample processing	How were samples collected, stored, prepared, and normalized? Was recovery measured?	Processing can change measured activity or the amount of relevant material in the assay.
Endpoint calculation	Was the reported value an observed rate, a fitted kinetic parameter, or a value converted to another basis?	Different calculations may produce different quantities from related experimental observations.
Population	Which donors, species, tissues, disease states, or other biological groups were represented?	Differences in group composition can change a study average even when group-specific values are similar.

The comparability matrix introduced in the earlier material on comparability is useful for organizing these checks. Here the emphasis is on explaining the discrepancy: identify a plausible source, verify it against reported methods or data, and state what remains unresolved. A plausible explanation is not proof. In particular, if assay design, biological material, and population all differ together, the published numbers alone may not distinguish their effects.

Endpoint and Assay Design

An important example arises when an assay measures reaction rate at a fixed substrate concentration but a paper reports a maximum rate estimated from several concentrations. Under the Michaelis–Menten model, the initial rate \(v\) at substrate concentration \(S\) is

v = \frac{V_{\max}S}{K_m+S}

Here \(V_{\max}\) is the maximum rate under the specified assay conditions, and \(K_m\) is the substrate concentration at which the model predicts \(v=V_{\max}/2\). This model is appropriate only when the initial-rate measurements and experimental system support its assumptions; the equation does not make unlike assays comparable automatically.

Worked Example: A Measured Rate Versus an Estimated Maximum Rate

Consider two hypothetical reports for the same enzyme preparation and assay system. Report A gives an observed initial rate of \(50\) pmol/min/mg protein at \(S=20\) μM. Report B fits initial rates over several substrate concentrations and estimates \(V_{\max}=100\) pmol/min/mg protein and \(K_m=20\) μM. Are the reported activity values necessarily contradictory?

Assume for this comparison that the initial-rate data follow the stated Michaelis–Menten model and that substrate concentration is the only relevant difference. At \(S=20\) μM, the model predicts

v = \frac{(100\ \text{pmol/min/mg})(20\ \text{μM})}{20\ \text{μM}+20\ \text{μM}} = \frac{2000}{40}\ \text{pmol/min/mg} = 50\ \text{pmol/min/mg}.

The predicted rate matches Report A. The numbers differ by a factor of two because one is a rate at a specified substrate concentration and the other is the estimated maximum rate. They are not estimates of the same endpoint, so their numerical difference does not establish conflicting measurements. If the assay conditions or biological material also differed, those would need separate checks.

This calculation illustrates why an assay description must be read alongside the reported result. A rate can also depend on incubation time, cofactor concentrations, probe-substrate selectivity, and how product formation is quantified. For example, a substrate used by more than one enzyme may produce a signal that is not specific to the enzyme of interest. Differences in analytical calibration or blank correction can affect reported rates as well. Check what was actually measured before interpreting a difference as a biological one.

Biological Material, Processing, and Normalization

Measurements from different material types may answer different questions. Recombinant enzyme can support estimation of activity for a defined enzyme system, whereas a tissue fraction contains an enzyme mixture and reflects its preparation. Activity per milligram of microsomal protein is not the same quantity as activity per gram of liver. A conversion between these bases requires a relevant measured yield and assumes that the activity and yield apply to the same material.

Worked Example: Checking a Tissue-Basis Conversion

Two hypothetical studies report activity in human liver microsomes. Study A reports \(2.4\) nmol/min/mg microsomal protein and a microsomal protein yield of \(35\) mg/g liver. Study B reports \(1.8\) nmol/min/mg microsomal protein and a yield of \(50\) mg/g liver. For illustration, calculate the implied activity per gram of liver for each study.

Assume each reported activity and its yield refer to the same sample basis, and that multiplying by the yield is an appropriate conversion without an additional recovery correction. For Study A,

2.4\ \frac{\text{nmol}}{\text{min}\cdot\text{mg}} \times 35\ \frac{\text{mg}}{\text{g liver}} = 84\ \frac{\text{nmol}}{\text{min}\cdot\text{g liver}}.

For Study B,

1.8\ \frac{\text{nmol}}{\text{min}\cdot\text{mg}} \times 50\ \frac{\text{mg}}{\text{g liver}} = 90\ \frac{\text{nmol}}{\text{min}\cdot\text{g liver}}.

On a microsomal-protein basis, Study B is \(0.6/2.4=0.25\), or 25%, lower than Study A. On the illustrative liver-tissue basis, the estimates are \(84\) and \(90\) nmol/min/g liver, a difference of \(6\) nmol/min/g liver. The estimates are closer on this converted basis, but the conversion does not show that either study is wrong or that the results are fully comparable. The assumed yields, recovery, and biological material must still be examined; this conversion alone does not establish a PBPK clearance parameter.

Sample processing can also affect what is recovered or remains active. Relevant details include time from collection to processing, storage temperature and duration, freeze–thaw history, homogenization, fractionation, and protein quantification. For instance, if activity is reported per milligram of recovered protein, a difference in recovery could influence the relationship between that denominator and the original tissue. Record the processing and denominator explicitly rather than assuming that a matching unit resolves the issue.

Population Composition and Study Averages

An average depends on the biological groups represented in the sample. Even if two studies measure the same endpoint using comparable methods, their means can differ because they include different proportions of groups with different activity. This is a population-composition explanation, not evidence that the assay itself is biased. It requires information about group-specific values and representation to evaluate.

Derivation: Suppose a study population contains a high-activity group with mean \(\mu_H\) and a low-activity group with mean \(\mu_L\). If the high-activity group makes up proportion \(w\), the population mean is \(\mu = w\mu_H+(1-w)\mu_L\). For two studies with proportions \(w_1\) and \(w_2\), but the same group-specific means, subtracting their means gives \(\mu_1-\mu_2=(w_1-w_2)(\mu_H-\mu_L)\). Thus, different group proportions alone can produce different overall means.

Worked Example: Different Group Proportions, Different Overall Means

Suppose two hypothetical studies use the same assay and normalization. Their represented high-activity group has a mean of \(140\) pmol/min/mg protein, and their low-activity group has a mean of \(60\) pmol/min/mg protein. Study A includes 70% high-activity and 30% low-activity samples; Study B includes 30% high-activity and 70% low-activity samples. Assume the group-specific means are applicable to both study samples and the reported average is a weighted arithmetic mean.

For Study A,

\mu_A = 0.70(140)+0.30(60) =98+18 =116\ \text{pmol/min/mg protein}.

For Study B,

\mu_B = 0.30(140)+0.70(60) =42+42 =84\ \text{pmol/min/mg protein}.

The difference is \(116-84=32\) pmol/min/mg protein. The group-specific means are identical in the example; the difference arises from group composition. In an actual comparison, this explanation would need support from the study populations and relevant subgroup data. If groups are defined differently, or the assay and processing differ, composition alone may not account for the discrepancy.

A Practical Diagnostic Sequence

Use a consistent order so that a numerical comparison does not outrun the evidence. First write each reported value with its units, endpoint, and denominator. Then compare methods, material and processing, calculation, and population. If a paper omits a detail, mark it as unknown rather than silently treating it as equal. The earlier material on measurement metadata and provenance supports this reconstruction.

Define the quantities.
State what activity is reported, its units, normalization basis, and whether it is a direct rate, fitted parameter, or converted value.

Compare assay conditions.
Check substrate, concentration, cofactors, incubation, analytical readout, and specificity. Identify conditions that could alter the measured rate.

Trace material and processing.
Record the biological source, preparation, storage, recovery, and denominator. Check whether any conversion uses a matching yield.

Examine population and calculation.
Check which groups were represented, how results were aggregated, and whether endpoint definitions or transformations differ.

State the diagnosis and its limits.
Separate documented differences from plausible explanations and unknowns. Do not claim a cause that the reported evidence cannot distinguish.

Comparisons, Diagnostics, and Common Pitfalls

Different approaches answer different diagnostic questions. A direct unit-and-endpoint comparison is quick, but it can miss differences hidden behind the same units. Reconstructing the assay and calculation is more informative when methods are reported in detail, but depends on complete reporting. A controlled bridge experiment, in which materials or methods are compared under shared conditions, can help isolate method effects; it may not be feasible and may not reproduce the original samples. Stratifying estimates by biological group can clarify composition effects when subgroup data exist, but cannot explain unmeasured differences. None of these approaches justifies pooling numbers merely because they are available.

Mistake: “The units match, so the estimates are comparable.” Matching units do not establish matching endpoints, assay conditions, denominators, or populations. Compare those features before interpreting the numerical gap.
Mistake: “The larger estimate must be the better measurement.” Magnitude alone does not establish assay validity or relevance to the intended biological quantity. Look for evidence about the method, material, calculation, and applicability.
Mistake: “A conversion makes the studies equivalent.” A conversion can change the basis of expression, but it relies on assumptions about yields, recovery, and matching samples. State those assumptions and retain the original reported value.
Mistake: “Different study means prove biological disagreement.” Different population composition, endpoints, or methods may explain the gap. Separate measured differences from conclusions about the underlying biology.

Diagnosis has limits. Published reports may omit assay details, sample-level observations, or subgroup composition. When several factors differ simultaneously, a plausible account may remain underdetermined. The appropriate conclusion is then specific about what is known, what could explain the discrepancy, and what evidence would be needed to distinguish explanations. This is more informative than averaging unlike measurements or treating uncertainty about comparability as evidence of agreement.

Check Your Understanding

For each item, identify what can be concluded and what additional information matters.

A study reports \(40\) pmol/min/mg protein at a fixed substrate concentration; another reports \(80\) pmol/min/mg protein as a fitted \(V_{\max}\). What should be checked before calling these conflicting estimates?
Two studies report \(3.0\) and \(2.5\) nmol/min/mg microsomal protein. Name two reasons matching units may still not establish comparability.
Using the group means \(140\) and \(60\) pmol/min/mg protein, calculate the mean if 50% of samples are in each group.
Why is multiplying activity per milligram of microsomal protein by a microsomal yield not automatically a valid conversion to activity per gram of liver?

Solutions. (1) Check whether both values are the same endpoint: the first is a rate at a stated substrate concentration, while the second is a fitted maximum. Compare substrate conditions and the assumptions and data supporting the fit. (2) Assay conditions, biological material, processing, endpoint calculation, normalization details, and population composition can differ despite matching units; any two are sufficient. (3) The weighted mean is \(0.50(140)+0.50(60)=70+30=100\) pmol/min/mg protein. (4) The conversion depends on a relevant yield for the same material and on assumptions about recovery and the relationship between the assayed fraction and tissue. Without those conditions, the product may not represent activity per gram of liver.