Correcting In Vitro Potency for Assay and Matrix Differences

Why the Same Compound Can Have Different In Vitro Potencies

A potency value is not a property of a compound in isolation. An IC50 or EC50 depends on what the assay measures, the assay conditions, and how much compound is available to interact with the target. For translation, a value measured in one species or laboratory may therefore be misleading when compared directly with a value from another assay. A defensible comparison first identifies the quantity measured and then corrects only for differences supported by the assay model.

Two distinctions are especially important. First, the concentration added to a well is usually a nominal total concentration; binding to proteins or other assay components can make the free concentration lower. Second, an observed IC50 may depend on a competitor or substrate concentration, whereas an affinity constant is intended to describe binding independently of that particular assay concentration. Neither correction, however, makes unlike functional systems automatically equivalent: receptor abundance and downstream signal amplification can change a functional EC50.

“Free” is relative to the experimental system: \(f_{u,\mathrm{assay}}\) in a cell-culture medium is not necessarily the plasma unbound fraction in an animal or patient. The correction is meaningful only when the fraction has been measured or credibly estimated in the same assay matrix, at relevant concentrations and conditions. A protein-containing medium, serum, cell lysate, plastic surface, or cell membrane can affect the available concentration in different ways.

Correcting a Competitive Assay for Substrate or Tracer

For a reversible competitive inhibitor of an enzyme, the measured IC50 depends on substrate concentration. Under the simple Michaelis–Menten model with competitive inhibition, the inhibition constant \(K_i\) is related to IC50 by the Cheng–Prusoff equation:

The correction follows from the rate equation. For competitive inhibition, the substrate-dependent rate is \(v=V_{\max}[S]/\{K_m(1+[I]_{\mathrm{free}}/K_i)+[S]\}\). Without inhibitor, \(v_0=V_{\max}[S]/(K_m+[S])\). At the IC50, inhibition halves the uninhibited rate, so \(v/v_0=1/2\). Substituting the two rates and cancelling common terms gives \(2(K_m+[S])=K_m(1+[I]_{\mathrm{free}}/K_i)+[S]\). Rearranging yields \([I]_{\mathrm{free}}=K_i(1+[S]/K_m)\), which is the stated equation. Thus, a higher substrate concentration produces a higher apparent IC50 even if the inhibitor’s \(K_i\) is unchanged.

A related correction applies to a simple equilibrium radioligand competition assay. If a test ligand competes reversibly with a tracer at one binding site, the tracer concentration is low enough to avoid depletion, and equilibrium is reached, then \(K_i=IC_{50}/(1+[L]/K_{d,L})\), where \([L]\) and \(K_{d,L}\) are the free tracer concentration and its equilibrium dissociation constant. This is not the enzyme-substrate equation by analogy alone; it follows from the competitive binding model. The quantities \(K_i\), \(K_m\), and \(K_{d,L}\) are not interchangeable.

The formula should not be applied automatically to every assay labelled “IC50.” It assumes the stated competitive mechanism and equilibrium model. Mixed or noncompetitive inhibition, irreversible inhibition, slow binding, substrate depletion, inhibitor depletion, or an uncertain \(K_m\) can make the result inappropriate. For receptor binding, the corresponding tracer correction likewise assumes a one-site competitive equilibrium model; a functional cellular IC50 is not automatically a binding \(K_i\).

Matrix Corrections and Comparison Across Assays

When the main difference is nonspecific binding in assay matrices, a first comparison is often the free concentration at the measured endpoint. If an assay reports nominal \(IC_{50}\), and \(f_{u,\mathrm{assay}}\) is known and approximately constant, then \(IC_{50,\mathrm{free}}=f_{u,\mathrm{assay}}IC_{50,\mathrm{nominal}}\). This places concentrations on a free-concentration scale. It does not correct differences in target sequence, assay endpoint, incubation time, cell permeability, or signalling pathway.

Matrix correction must be based on the matrix actually used. A plasma unbound fraction should not be substituted for the unbound fraction in a cell assay. If binding is concentration-dependent, one constant multiplier may also be wrong: the appropriate free concentration must be estimated at each relevant nominal concentration, or the concentration-response analysis should incorporate the concentration-dependent binding relationship. Similarly, adsorption or precipitation can reduce available compound but should not be labelled protein binding unless that process has been measured and modelled appropriately.

Receptor Expression and Functional Potency

Receptor expression is a distinct issue from free concentration. For a simple equilibrium interaction, the fraction of receptors occupied by a ligand is \(\theta=C_{\mathrm{free}}/(K_d+C_{\mathrm{free}})\), where \(K_d\) is the ligand’s dissociation constant. At a fixed free concentration and fixed \(K_d\), changing receptor abundance does not change this fractional occupancy. It can, however, change the measured response: more receptors can provide greater signal amplification or receptor reserve, so a given occupancy may produce a larger response. Consequently, a functional EC50 can shift when receptor expression changes even if affinity does not.

There is no general multiplier based only on the ratio of receptor expression values that converts one functional EC50 into another. Such a correction would require a justified model linking receptor abundance, occupancy, and response, supported by measurements. When suitable binding or target-engagement data are available, a more interpretable comparison is often the free concentration required for a specified occupancy. That comparison describes target engagement, not necessarily equal functional effect.

Assumptions, Diagnostics, and Common Pitfalls

Before comparing values, identify whether each result is a binding IC50, a functional IC50 or EC50, or a fitted affinity constant. Record assay matrix, free fraction, substrate or tracer concentration, relevant \(K_m\) or \(K_d\), incubation time, temperature, pH, and receptor expression. These are not administrative details: they determine whether a correction equation applies and whether the corrected measurements remain comparable.

Useful warning signs include IC50 estimates that change markedly with assay duration, poor agreement between replicate concentration-response curves, unexpected shifts with substrate concentration, or apparent potency that depends on cell density. These can indicate failure to reach equilibrium, substrate or ligand depletion, concentration-dependent binding, aggregation, assay interference, or changes in receptor availability. A correction based on a single \(f_u\), \(K_m\), or tracer \(K_d\) is not reliable if that quantity is uncertain or changes materially across the fitted concentration range.

A common mistake is to treat every measured IC50 as an intrinsic affinity and “correct” it by multiplying by a plasma protein-binding value. This combines two errors: IC50 is assay-condition dependent, and plasma binding is not necessarily the binding in the assay well. The correction is to use the actual assay matrix and endpoint, convert to free concentration only with an assay-relevant unbound fraction, and apply a competitive-model correction only when its assumptions hold. If receptor expression or functional signalling differs, report that limitation rather than inventing a universal numerical adjustment.