Why Assay Metadata Must Be Reconstructed
A published enzyme-activity value is not fully described by its number and units. Its interpretation depends on what assay generated it, which biological material was tested, what was measured, under what protocol and conditions, how many independent experiments contributed, and what statistic summarizes those experiments. Reconstructing these details creates a usable record of the measurement and its provenance; it does not, by itself, establish that the measurement is suitable as a PBPK input.
The central task is to translate methods text, tables, figure legends, and supplements into discrete fields while preserving what the publication actually says. A method can be spread across several locations, and a detail that seems routine may still be unreported. The disciplined response to an unreported detail is to mark it as not stated or unresolved, not to supply a conventional value from memory.
Fields to Reconstruct
Keep the fields separate even when a sentence supplies information for more than one of them. “Human liver microsomes were incubated with substrate at 37°C and product was quantified by LC–MS/MS” identifies biological material, part of the protocol, a condition, and an analytical method. It does not necessarily specify the full assay type, the number of independent experiments, or the statistic reported.
| Field | What to record | Question to ask |
|---|---|---|
| Assay type | The measurement approach, such as enzyme activity measured by product formation or substrate depletion; note if the readout is direct or coupled when stated. | What kind of assay produced the result? |
| Biological material | The biological source and preparation, such as recombinant enzyme, microsomes, or a named cell preparation; include species or tissue when given. | What was tested? |
| Protocol | The operational steps needed to understand the measurement, including reaction setup, duration, quenching, and analytical readout when reported. | How was the measurement performed? |
| Analyte | The quantity measured or quantified, such as a named product, substrate, or signal. Distinguish it from the enzyme under study. | What was measured? |
| Conditions | Reported experimental settings such as temperature, pH, substrate concentration, incubation time, and cofactors. | Under what stated conditions? |
| Replicate count | The number of independent biological preparations or experiments and the number of technical repeats, if both are stated. | What does the reported \(n\) count? |
| Reported statistic | The summary form, such as mean with standard deviation, and the observations to which it applies if specified. | How are results summarized? |
The biological material is not interchangeable with the analyte. In an enzyme-activity experiment, the enzyme may be the biological target, while the analyte is the product formed or substrate remaining. Likewise, an instrument signal is not automatically the analyte: it may be a signal used to quantify the analyte through calibration or a coupled reaction. Record the publication’s stated measurement chain rather than collapsing these distinctions.
Use a source locator for each entry, such as the methods subsection, figure legend, table, or supplement section. A concise quotation or faithful paraphrase can help preserve the basis for the entry. Mark each item as explicitly stated, calculated from explicitly stated values, or unresolved. These labels prevent an inference from later appearing to be a directly reported fact.
A Worked Enzyme-Activity Record
Worked Example: Reconstructing a CYP3A4 Activity Assay
Hypothetical publication text. The methods state: “Recombinant human CYP3A4 microsomes were incubated with testosterone and NADPH for 10 min at 37°C in pH 7.4 buffer. Formation of 6β-hydroxytestosterone was quantified by LC–MS/MS. Initial product-formation rates were determined from the linear interval. Three independent incubations were performed, each in triplicate wells. The mean and standard deviation across independent incubations are reported.” A supplement gives the three normalized rates as 19, 21, and 23 nmol min\(^{-1}\) mg\(^{-1}\) protein.
Assay type. This is an in vitro enzyme-activity assay based on product formation, with LC–MS/MS as the stated analytical readout. The methods identify a quantified product; they do not describe a coupled signal, so none should be added to the record.
Biological material. Record recombinant human CYP3A4 microsomes. Do not shorten this to “human liver microsomes”: the text describes recombinant enzyme in microsomes, not a liver-tissue preparation.
Analyte. The quantified analyte is 6β-hydroxytestosterone. CYP3A4 is the enzyme under study, and testosterone is the substrate; neither is the measured product named in the assay description.
Protocol and conditions. Record incubation with testosterone and NADPH, followed by quantification of product by LC–MS/MS. The explicitly stated conditions are 10 min, 37°C, and pH 7.4 buffer. “Initial product-formation rates” describes the rate calculation, but the text does not give a substrate concentration, buffer composition, or exact linear-interval duration. Those entries remain unreported.
Replicates and statistic. The biological replicate count is three independent incubations. Each had three technical wells, so there were nine wells in total, but not nine independent incubations. The stated summary is mean and standard deviation across the three independent incubation rates.
Check the arithmetic. The mean rate is
Using the sample standard deviation \(s\), the deviations from the mean are \(-2\), \(0\), and \(2\) nmol min\(^{-1}\) mg\(^{-1}\) protein. Therefore,
The reconstructed reported statistic is thus \(21 \pm 2.0\) nmol min\(^{-1}\) mg\(^{-1}\) protein, with the plus-minus notation explicitly identified as mean ± sample standard deviation across three independent incubations. This calculation does not imply that the publication reported an uncertainty interval.
How to Reconstruct the Record Reliably
A reliable record is built by searching for the experimental description and then checking where the results are defined. Methods commonly describe materials and procedures; figure legends may specify the plotted statistic and \(n\); supplements may provide concentrations, replicate structure, or additional protocol detail. Do not assume one section contains the complete account.
Find the assay-specific methods and any linked figure, table, or supplement. Record the source location before summarizing the method.
Identify assay type, biological material, protocol, analyte, and conditions independently. If a sentence supports several fields, use it as evidence for each relevant entry without merging them.
Find whether \(n\) refers to independent preparations, experiments, animals, wells, or repeated measurements. Record biological and technical counts separately when specified.
Check whether the result is a mean, median, range, standard deviation, standard error, or another summary. Record what the publication says the statistic summarizes.
Use “not stated” or “unresolved” for missing details. Preserve a brief note describing the ambiguity instead of choosing the most familiar interpretation.
The logic behind the replicate check is important: a statistic describes the observations included in its calculation, and the independent experimental unit determines what those observations represent. If three independently prepared samples are each measured in triplicate, the nine well readings may be technical repeats nested within three preparations. Treating all nine wells as independent would change the implied sample size and could misrepresent the experiment. If the article does not say whether wells were averaged first, do not claim that it did.
Similarly, reported activity units should be preserved as published before any later transformation. If a rate is derived from measured product, the dimensional relationship can be checked. For example, product amount divided by time gives an amount-per-time rate; dividing that rate by protein mass gives a mass-normalized activity. This check can reveal a unit mismatch, but it cannot supply missing protocol facts or show that two different assays are equivalent.
Worked Example: Separating Biological and Technical Replicates
Hypothetical publication text. A figure legend says “data are mean ± SD, \(n=6\).” The methods state that liver microsomes from three individual donors were tested, with duplicate wells for each donor, but do not explicitly say whether the duplicate wells were averaged before calculating the plotted statistic.
Assay metadata that can be recorded. The biological material is liver microsomes from three individual donors. The reported design includes duplicate wells per donor: two technical measurements for each of three biological sources. There are six well-level measurements, but only three donor sources.
What remains unresolved. The figure legend’s \(n=6\) could refer to six wells, while the methods identify three donors. The publication text provided here does not establish whether the plotted mean and SD were calculated across six well readings or across three donor-level values after averaging duplicate wells. Record “\(n=6\), definition ambiguous; three donors, duplicate wells per donor; aggregation before summary not stated.” Do not silently replace the reported \(n=6\) with \(n=3\), and do not describe six wells as six independent donors.
This reconstruction is useful precisely because it preserves both the stated count and the uncertainty about what was summarized. It does not make an unsupported choice about the calculation.
Worked Example: Using a Supplement to Complete Protocol Metadata
Hypothetical publication text. The main methods say that an enzyme preparation was incubated with substrate and that product formation was measured. The supplement specifies a 15-min incubation at 30°C, an NADPH-containing buffer at pH 7.5, and quantification of a named hydroxylated product by LC–MS/MS. The figure legend reports “mean ± SEM, \(n=4\),” and the supplement says that each of four independent preparations was measured once.
Protocol and conditions. Combine the main methods and supplement into a single protocol summary, retaining the source location for the added details. Record the 15-min incubation, 30°C, pH 7.5, and NADPH-containing buffer as stated conditions. Record LC–MS/MS as the product-quantification method.
Analyte and assay type. The analyte is the named hydroxylated product, not the enzyme or substrate. The assay type is an enzyme-activity measurement based on product formation, with LC–MS/MS detection.
Replicates and statistic. The reported \(n=4\) corresponds to four independent preparations, each measured once; no technical replicate count beyond one is reported. The statistic is mean ± standard error of the mean (SEM), not mean ± standard deviation. Since no individual rates or SD are provided in this hypothetical text, the SEM cannot be independently recalculated from the available information. Record the statistic as reported rather than trying to reconstruct a missing SD.
Diagnostics, Alternatives, and Limits
Several warning signs indicate that a metadata record needs further checking. A result with units but no named analyte may have been copied from a table without its caption. A value labeled “activity” may refer to a signal, a product-formation rate, or a normalized rate; the methods should clarify which. An \(n\) in a figure legend may count wells, preparations, donors, or experiments. A protocol may mention a measured product but omit the analytical method. Each is a documentation gap to flag, not a license to fill in a conventional detail.
The main practical alternative to a structured record is a single prose summary or direct copying of a table entry. Prose can be concise, but it can hide omissions and blur distinctions such as enzyme versus analyte. Copying a value preserves the number but may detach it from its replicate definition, units, and assay conditions. A field-based record makes those elements visible and easier to audit. It is not a substitute for the publication itself: retain source locations so the underlying wording can be checked.
Metadata reconstruction also has limits. Publications may omit details, use inconsistent labels, or describe a procedure across multiple documents. Sometimes a supplement clarifies one field but leaves another ambiguous. The record should state what is known and what is not, rather than presenting a complete-looking assay description assembled from assumptions. Nor does a complete metadata record prove that the result is transferable to a model context; judging whether measurements are comparable or selecting an input requires additional evaluation.
Common Mistakes
- Wrong reasoning: “The paper says \(n=9\), and there were three wells per sample, so there must have been nine independent biological replicates.” Correction: wells may be technical repeats. Identify the independent preparations or subjects separately and record the publication’s aggregation rule if given.
- Wrong reasoning: “CYP3A4 is the analyte because the experiment measures CYP3A4 activity.” Correction: CYP3A4 is the enzyme under study; the analyte may be the product or substrate quantified by the assay.
- Wrong reasoning: “A standard assay would use 37°C, so that is the condition even though the paper does not say.” Correction: record only stated conditions and label missing conditions as unreported.
- Wrong reasoning: “Mean ± SEM and mean ± SD both describe spread, so either label is adequate.” Correction: they are different statistics. Preserve the reported statistic and its stated basis.
- Wrong reasoning: “The methods are incomplete, so I can fill in a likely protocol from another paper.” Correction: keep the record specific to the source being documented; mark the gap rather than importing undocumented details.
Check Your Understanding
Answer each question, then compare your reasoning with the solutions below.
- A methods section names an enzyme and a product quantified by mass spectrometry. Which is the analyte, and what other role does the enzyme have?
- A study reports three independent preparations, each measured in duplicate wells. How many biological replicates and how many technical wells are described?
- A legend reports “mean ± SEM, \(n=5\),” but does not define \(n\). What should the metadata record say about replicate count?
- A protocol states the temperature and incubation duration but not the substrate concentration. Should a familiar concentration be entered? Explain.
- Why should a record preserve whether a reported “±” value is SD or SEM?
Solutions. (1) The quantified product is the analyte; the enzyme is the biological target or catalyst whose activity is being assessed. (2) There are three biological replicates and six technical wells in total, with two wells per preparation. (3) Record mean ± SEM as the reported statistic and \(n=5\) as stated, but mark what \(n\) counts as unresolved unless another source passage defines it. (4) No. Record the stated temperature and duration and mark substrate concentration as not stated. (5) SD and SEM represent different summaries: SD describes sample dispersion, whereas SEM describes the estimated standard error of the sample mean. Substituting one label for the other changes the meaning of the reported result.