Documenting Transformations From Measurement to Input

Why the Transformation Record Matters

A PBPK model input is often not the number printed in a paper. A reported measurement may need a unit conversion, normalization to an assay denominator, a correction for incubation binding, or scaling from microsomal protein to liver. Each operation can be reasonable, but an undocumented operation makes it difficult to reproduce the input or determine what quantity the model actually uses.

A useful record therefore preserves the reported value as reported and separately documents each transformation. It should make clear which quantities came from the source, which were calculated, which were selected from another source, and what the final value represents. The earlier material on parameter records, measurement comparability, reported qualifiers, and tracing secondary citations supplies the provenance context; the focus here is the arithmetic and its documentation.

State the Quantity Before Calculating

The same label can refer to different quantities. For microsomal clearance, distinguish at least: a depletion rate measured in an incubation; clearance normalized per milligram of microsomal protein; an apparent or binding-corrected in vitro intrinsic clearance; and a value scaled to a liver or body-weight basis. Before calculating, state which of these is the target. Otherwise, a numerically plausible result can still be the wrong input.

For the examples below, the scaling chain assumes that the assay value is appropriate for the intended compound and biological system; the protein amount in the incubation is known; and the selected microsomal protein per gram of liver (MPPGL) and liver mass refer to the species and scaling basis being modeled. Whole-liver scaling also assumes that the measured microsomal activity can be extrapolated across the liver using those factors. These are assumptions to record, not consequences of unit conversion.

An incubation-binding correction has an additional assumption: the unbound fraction in the incubation, \(f_{u,\mathrm{inc}}\), is relevant to the measured assay conditions, and unbound drug drives the measured metabolic clearance. Under that interpretation, the unbound intrinsic clearance is the apparent intrinsic clearance divided by \(f_{u,\mathrm{inc}}\). If the model instead requires the apparent value, or already handles binding elsewhere, applying this correction may be inappropriate or may count the same effect twice.

From Incubation Depletion to a Scaled Input

One common route starts with a first-order depletion estimate. If the measured concentration declines with rate constant \(k\) in an incubation of volume \(V_{\mathrm{inc}}\), and the incubation contains \(m_{\mathrm{protein}}\) milligrams of microsomal protein, apparent clearance per milligram of protein is calculated as \(kV_{\mathrm{inc}}/m_{\mathrm{protein}}\). This calculation presumes the depletion estimate is suitable for a first-order clearance summary over the interval used, and that the stated volume and protein amount match that assay.

If \(k\) is in \(\mathrm{min}^{-1}\), \(V_{\mathrm{inc}}\) in mL, and protein amount in mg, the result is mL/min/mg. Converting mL to μL gives the commonly reported μL/min/mg. If a binding correction is justified, \(CL_{\mathrm{int,u}}=CL_{\mathrm{int,app}}/f_{u,\mathrm{inc}}\). Scaling to a whole liver then multiplies the per-milligram value by MPPGL and liver mass. The units must cancel to a volume per time.

This expression yields μL/min if \(CL_{\mathrm{int,u}}\) is in μL/min/mg. Convert μL to mL or L only after multiplication, or explicitly within the calculation. For a body-weight-normalized result, divide whole-liver clearance by body weight, giving units such as mL/min/kg.

A Short Unit-Cancellation Check

The scaling calculation is justified by its units as well as by its biological assumptions. Clearance per milligram of microsomal protein is multiplied by milligrams of microsomal protein per gram of liver, then by grams of liver. The mg and g units cancel, leaving volume per time:

If the result does not have the intended units, the chain contains a missing, inverted, or duplicated factor. Passing this dimensional check does not prove that the chosen scaling factors apply biologically; it confirms only that the arithmetic is dimensionally coherent.

What a Reproducible Record Contains

Keep the chain in a form another analyst can recalculate. The record need not be elaborate, but it should distinguish source facts from derived quantities and expose every factor. A compact table can make the sequence auditable.

The record should also retain enough precision in intermediate calculations to reproduce the displayed result. Report a sensible number of significant figures in the final value, but avoid rounding at each step if that would materially alter the result. A stated correction must have a stated purpose: for example, “divide by \(f_{u,\mathrm{inc}}\) to obtain the unbound intrinsic clearance,” not simply “corrected.”

Diagnostics, Alternatives, and Common Errors

Several checks help detect a failing transformation chain. Confirm that each factor has compatible species, tissue, and assay context; verify that a concentration has not been mistaken for an amount; and cancel units at every multiplication or division. Recalculate the result from the retained values rather than copying a final number. If the result changes by exactly 1000-fold when switching between μL and mL, inspect whether the conversion was applied once, in the right direction.

The main alternative to microsomal scaling is to use a more directly measured whole-liver or in vivo quantity when that quantity is appropriate for the model and evidence question. Direct measurements avoid some microsomal extrapolation factors, but may represent a different endpoint or depend on additional assumptions. A microsomal value offers an assay-based route to a mechanistic input, but requires normalization and scaling assumptions. These approaches are not interchangeable merely because their final units match.

Similarly, apparent and unbound intrinsic clearance are alternatives only when the model formulation calls for one or the other. Applying \(1/f_{u,\mathrm{inc}}\) to an apparent value and then later applying an equivalent correction again is double correction. Conversely, omitting the correction when the intended input is unbound clearance may leave the model with a different quantity than intended. Record the target definition and where any binding treatment occurs.

• Wrong reasoning: “The paper reports 12, so I entered 12.” Correction: Preserve 12 with its original units and state whether the model needs a normalized, corrected, or scaled value.
• Wrong reasoning: “The final units look right, so the calculation must be right.” Correction: Unit consistency is necessary but does not validate species matching, correction assumptions, or scaling applicability.
• Wrong reasoning: “I converted the liver mass to kilograms because body weight is in kilograms.” Correction: Use the units required by the factor being multiplied; MPPGL in mg/g requires liver mass in g.
• Wrong reasoning: “The binding correction is always required for microsomal clearance.” Correction: Apply it only when the target is defined as unbound clearance and the assay/model assumptions support it; document where binding is handled.

Microsomal scaling is limited by the validity of the assay estimate, the representativeness of the binding measurement, and the applicability of MPPGL and liver mass to the intended system. A transparent record does not remove those limitations; it makes the assumptions and calculations inspectable. It also makes later review of uncertainty and applicability limits possible without obscuring what value was originally reported.