Tutorials › Model-Informed Drug Development (MIDD) › Deriving MABEL from In Vitro Potency and Cellular Assays

Home  /  Tutorials  /  First-in-Human Dose Selection

First-in-Human Dose Selection · Tutorial 89 of 596

Deriving MABEL from In Vitro Potency and Cellular Assays

Use functional concentration–response data, supported by binding measurements, to derive a defensible minimal anticipated biological effect level and understand the assumptions behind its dose estimate.

Intermediate 11 min read

What You'll Learn

  • Distinguish binding potency from functional potency and explain why they are not interchangeable.
  • Invert a Hill concentration–response model to calculate a concentration for a specified minimal response.
  • Combine assays by considering biological relevance, assay quality, and uncertainty rather than selecting the smallest reported potency value.
  • Correct nominal assay concentrations for unbound fractions in assay media and human plasma.
  • Translate an exposure target into a provisional dose using an explicit pharmacokinetic relationship.
  • Recognize assay and translation assumptions that can make a MABEL estimate unreliable.

From In Vitro Potency to a Minimal-Effect Concentration

A first-in-human starting-dose assessment may need to anticipate biological activity at exposures below those tested in animals. In vitro binding and cellular assays can help identify such exposures, but a binding measurement alone does not say how much functional response a human exposure will produce. The central task is to connect evidence about target interaction to a relevant functional response, then estimate the concentration associated with a small, specified effect.

A MABEL derived from these data is an estimate, not a directly observed human threshold. Its strength depends on whether the assays represent the human pharmacology of interest, whether their concentration–response results are interpretable, and whether the in vitro concentration can be related to the relevant unbound concentration in humans. The introductory material on MABEL from mechanistic evidence provides the broader role of this estimate; here the focus is deriving it from potency and cellular-assay evidence.

Definition: For an in vitro potency and cellular-assay analysis, the MABEL concentration is an estimated drug concentration associated with a small, prespecified functional response in a biologically relevant assay. The threshold response is a modeling choice guided by the assay and context; 10% of the assay’s maximal response is a useful illustrative convention, not a universal regulatory threshold. A MABEL-based dose is a provisional dose predicted to produce the corresponding target exposure, subject to the assumptions used for translation.

Use Functional Activity to Set the Target

A binding assay measures an interaction such as ligand binding or competition at a target. Results may be reported as a dissociation constant, \(K_d\), or an assay-specific \(IC_{50}\). A cellular assay instead measures a functional consequence, such as signaling, cytokine release, or inhibition of a response. Its potency may be summarized by \(EC_{50}\) for stimulation or \(IC_{50}\) for inhibition. These values answer different questions: binding supports the proposed mechanism, while a relevant functional assay estimates the concentration–response relationship used to set a minimal-effect target.

For a monotone stimulatory response, a common model is the Hill concentration–response model. Let \(C\) be the concentration used in the assay, \(E_0\) the baseline response, \(E_{\max}\) the maximal change from baseline, \(EC_{50}\) the concentration producing half of that change, and \(h\) the Hill coefficient:

$$ E(C)=E_0+E_{\max}\frac{C^h}{EC_{50}^h+C^h}, \qquad C\geq 0,\quad EC_{50}>0,\quad h>0. $$

The fraction of maximal response above baseline is \(r=(E(C)-E_0)/E_{\max}\). For a target fraction \(0<r<1\), solving the model for \(C\) gives:

$$ r=\frac{C^h}{EC_{50}^h+C^h} \quad\Longrightarrow\quad C_{\text{target}}=EC_{50}\left(\frac{r}{1-r}\right)^{1/h}. $$

This inversion assumes the response follows the fitted Hill model over the relevant concentration range, the baseline and maximal response are meaningfully estimated, and the response is monotone. If the assay is inhibitory, define the modeled fractional inhibition above baseline and use its \(IC_{50}\) in the same expression. A reported \(IC_{50}\) should not be treated as a universal molecular constant: it depends on the assay design and conditions.

Worked Example: Estimate a Ten-Percent Functional Effect

In a hypothetical human-cell assay, a drug produces a stimulatory response with \(EC_{50}=2.00\) nM and fitted Hill coefficient \(h=1.30\). Assume the baseline-adjusted response follows the Hill model and define the illustrative minimal response as 10% of the fitted maximum. Estimate the corresponding assay concentration.

Here \(r=0.10\), so:

$$ C_{10}=2.00\left(\frac{0.10}{1-0.10}\right)^{1/1.30} =2.00\left(\frac{1}{9}\right)^{0.7692} \approx 2.00(0.1845) =0.369\ \text{nM}. $$

Under the fitted model, 0.369 nM produces 10% of the maximal response above baseline. This is a model-based estimate for this assay, not evidence that the same concentration produces exactly 10% activity in every human tissue.

Combine Assays Without Treating Potencies as Interchangeable

The most useful evidence set usually includes a target-binding or competition assay and one or more functional assays. Binding potency can support target engagement and help interpret whether functional potency is plausible. It does not, by itself, establish the functional response at a given concentration. Signal amplification, receptor abundance, coupling efficiency, and assay conditions can all separate binding potency from functional potency.

Prioritize functional data from human cells and pathways relevant to the intended pharmacology. Consider whether the assay has a suitable positive control, a stable baseline, adequate concentration coverage, and reproducible responses. If several relevant assays differ, do not simply average their \(EC_{50}\) or \(IC_{50}\) values: each assay may represent a different biological system or endpoint. Explain which evidence is most relevant and retain meaningful uncertainty. A lower concentration from a reliable, relevant assay may warrant a more conservative target; a low value from an overexpressing or otherwise poorly representative system needs interpretation rather than automatic adoption or dismissal.

Assay concentrations may be nominal rather than unbound. If the measured unbound fraction in assay medium is \(f_{u,\text{assay}}\), and binding is adequately characterized over the concentration range, the estimated unbound potency concentration is:

$$ C_{u,\text{assay}}=f_{u,\text{assay}}C_{\text{nominal}}. $$

If unbound drug is assumed to drive the same functional effect in humans, the corresponding total plasma concentration target is \(C_{\text{plasma,target}}=C_{u,\text{target}}/f_{u,\text{plasma}}\), where \(f_{u,\text{plasma}}\) is the human plasma unbound fraction. This translation assumes that the relevant unbound concentration is comparable between assay and human tissue, and that plasma unbound concentration is an appropriate exposure surrogate. Those assumptions may be weak for drugs with substantial tissue partitioning, active metabolites, or concentration-dependent binding.

Worked Example: Correct for Unbound Drug and Estimate a Dose

Suppose a hypothetical functional assay has a nominal \(EC_{50}=1.50\) nM and \(h=1.00\). The measured unbound fraction in assay medium is 0.80. Human plasma \(f_{u,\text{plasma}}\) is estimated to be 0.050. A preliminary linear pharmacokinetic projection predicts a total-plasma \(C_{\max}\) of 0.200 nM per mg administered. Estimate the dose corresponding to a 10% maximal response, assuming unbound concentration drives the effect.

First convert the nominal \(EC_{50}\) to an unbound assay concentration: \(1.50\times0.80=1.20\) nM. With \(h=1\), the 10% response concentration is one ninth of \(EC_{50}\):

$$ C_{u,\text{target}}=1.20\left(\frac{0.10}{0.90}\right) =0.133\ \text{nM}. $$

Convert this to a total-plasma target using the stated human unbound fraction:

$$ C_{\text{plasma,target}}=\frac{0.133}{0.050} =2.66\ \text{nM}. $$

Finally, use the projected proportionality between dose and total-plasma \(C_{\max}\):

$$ \text{Dose}=\frac{2.66\ \text{nM}}{0.200\ \text{nM/mg}} =13.3\ \text{mg}. $$

The 13.3 mg result is a provisional dose estimate under the specified assumptions, not a recommended clinical dose by itself. It assumes linear pharmacokinetics in this range, a valid \(C_{\max}\)-per-dose projection, and that matching the unbound concentration target is meaningful for the functional effect.

Assumptions, Diagnostics, and Failure Modes

Before relying on an estimate, check that the assay data support the fitted concentration–response model. The tested range should include enough information to locate the baseline, response rise, and (where feasible) upper plateau; otherwise \(EC_{50}\), \(E_{\max}\), or \(h\) may be poorly identified. Inspect replicate variability, control responses, nonspecific effects, and whether the response is reproducible across donors or cell preparations. A fitted \(EC_{50}\) extrapolated far beyond the concentrations tested should not be presented as a precise measurement.

The Hill inversion also presumes the specified fractional response is meaningful. If the assay has a noisy or drifting baseline, a small fraction of the fitted maximum may be indistinguishable from assay variability. If the response is nonmonotone, has multiple phases, or is driven by cytotoxicity at higher concentrations, one Hill curve may be misleading. Use an appropriate alternative model or report that the assay does not support a reliable threshold estimate; do not force a precise MABEL concentration from an unsuitable fit.

Uncertainty can enter through assay potency, Hill slope, unbound fractions, assay relevance, and human exposure prediction. Assess how the derived concentration changes across plausible parameter values and assay choices. If a small change in \(h\) or \(EC_{50}\) substantially changes the target, the estimate is sensitive and should be treated accordingly. The dose conversion adds uncertainty from human pharmacokinetics and the assumed exposure metric. It is not justified to report more precision in the final dose than the evidence supports.

A direct concentration–response estimate from a relevant functional assay is generally more informative for minimal activity than using a binding \(K_d\) alone. Binding data are useful corroboration, but converting \(K_d\) directly into a functional threshold would require additional assumptions about target abundance, interaction, and signaling. Animal NOAEL- or HNSTD-based approaches answer a different question about observed toxicology and do not substitute for estimating low-level human pharmacology. These evidence streams can inform the overall starting-dose assessment, but each should be interpreted according to what it measures.

Common pitfall: “The binding \(K_d\) is 0.1 nM, so the MABEL concentration must be 0.1 nM.” This treats binding and functional response as equivalent. Instead, use a relevant cellular concentration–response relationship to estimate the specified small effect, use binding results as supporting mechanistic evidence, and state the assumptions needed to translate assay concentration into human exposure.

A Practical Derivation Sequence

A defensible derivation is easiest to review when each transition is explicit. The sequence below separates assay interpretation from exposure and dose translation.

1
Select interpretable evidence.
Identify relevant human functional assays and supporting binding measurements; document assay conditions, concentration ranges, and response quality.
2
Define the minimal response.
Specify the functional endpoint and response fraction to target. Explain why that threshold is suitable for the evidence rather than implying that 10% is universal.
3
Estimate the assay concentration.
Fit or otherwise characterize the concentration–response relationship, then calculate the concentration associated with the target response and assess its uncertainty.
4
Translate concentration and dose.
Correct for relevant unbound fractions and apply an explicit human exposure–dose projection, stating the assumptions and limitations at each step.

Check Your Understanding

Answer each question, then compare with the solutions below.

  1. A functional assay has \(EC_{50}=4.0\) nM and \(h=1\). What concentration corresponds to 10% of maximal response?
  2. Why should a binding \(K_d\) not automatically be used as a functional MABEL concentration?
  3. An assay gives a nominal target concentration of 2.0 nM and \(f_{u,\text{assay}}=0.50\). What is the estimated unbound assay concentration?
  4. Name two diagnostics that would make a Hill-model threshold estimate questionable.

Solutions. (1) \(C_{10}=4.0(0.10/0.90)=0.444\) nM. (2) Binding measures target interaction, whereas functional potency depends on the cellular response and assay context; the two are not interchangeable. (3) \(2.0\times0.50=1.0\) nM unbound. (4) Examples include a tested concentration range that does not define the response curve, an unstable baseline, poor reproducibility, nonmonotone response, or a threshold extrapolated far outside the data.

Previous

Next

Back to Tutorials