Interpreting Selectivity Across Multiple Targets

Why Selectivity Depends on Concentration

A compound can bind one target more tightly than another and still occupy both targets substantially at sufficiently high free concentration. Conversely, at very low concentration, occupancy of both targets may be small even when their affinities differ greatly. Thus, an affinity comparison and an occupancy comparison answer related but distinct questions: affinity describes binding strength, while occupancy describes the predicted fraction of a target bound at a specified concentration.

To make a meaningful occupancy comparison, evaluate each target at the same free drug concentration. Comparing occupancies at different concentrations can make a target appear more or less selective simply because the exposure differs. The comparison here uses a simple equilibrium model; it does not describe competition among ligands or more complex binding mechanisms.

This model assumes a single class of independent binding sites, reversible binding at equilibrium, and a known free concentration that is effectively the same for both targets. It also assumes that binding does not appreciably deplete free drug and that the reported \(K_d\) values refer to comparable conditions. Under these assumptions, lower \(K_d\) means higher affinity, but occupancy also depends on \(C\).

Affinity Ratios and Occupancy Comparisons

A useful way to separate affinity selectivity from fractional occupancy is to compare the occupancy odds, \(\theta_j/(1-\theta_j)\). Rearranging the one-site equation gives \(\theta_j/(1-\theta_j)=C/K_{d,j}\). Therefore, when both targets are evaluated at the same positive free concentration,

The ratio of occupancy odds equals the affinity selectivity ratio and does not vary with concentration in this model. The ratio of fractional occupancies, \(\theta_A/\theta_B\), does vary with concentration, as does their difference, \(\theta_A-\theta_B\). At low concentrations, occupancy is approximately proportional to concentration divided by \(K_d\), so the fractional-occupancy ratio approaches the affinity ratio. At high concentrations, both occupancies approach one, so their ratio approaches one and their difference approaches zero. These patterns explain why a strong affinity difference need not look like a similarly large difference in fractional occupancy at every concentration.

Where Is the Occupancy Difference Largest?

For two targets where \(K_{d,A}<K_{d,B}\), the occupancy difference at a common concentration is

This difference is zero at \(C=0\), and tends toward zero as \(C\) becomes very large. Its maximum occurs between those extremes. Differentiating with respect to \(C\), and writing \(a=K_{d,A}\) and \(b=K_{d,B}\), gives

Because \(b-a>0\) and the denominator is positive, the derivative is positive when \(C^2<ab\), zero when \(C^2=ab\), and negative when \(C^2>ab\). Consequently, the maximum occurs at \(C=\sqrt{K_{d,A}K_{d,B}}\), the geometric mean of the two dissociation constants. Substitution into \(\Delta(C)\) gives

This result concerns the absolute difference in fractional occupancy, measured in percentage points. It does not mean that the drug is selectively binding only one target at that concentration; it identifies where the model predicts the largest separation in occupancy between the two targets.

Using Occupancy Thresholds to Assess a Selective Range

A practical question is whether any concentration can achieve a specified occupancy on a desired target while keeping occupancy of another target below a specified limit. For example, under the one-site model, \(\theta_A\geq0.90\) requires \(C\geq9K_{d,A}\): this follows by solving \(C/(K_{d,A}+C)\geq0.90\). Likewise, \(\theta_B\leq0.10\) requires \(C\leq K_{d,B}/9\). Both conditions can hold only if \(9K_{d,A}\leq K_{d,B}/9\), or \(K_{d,B}/K_{d,A}\geq81\).

Assumptions, Diagnostics, and Alternatives

The equations are only as useful as their assumptions and inputs. A selectivity comparison should state the targets, the \(K_d\) values and their measurement conditions, the matched free concentration, and whether occupancy is measured or model-predicted. Check that concentrations and \(K_d\) values use the same units. If total concentration is substituted for free concentration, the calculation can be misleading when protein binding, nonspecific binding, or target binding makes the free fraction materially different or target-dependent.

Look for signs that the simple model is not adequate. Occupancy observations that systematically depart from \(C/(K_d+C)\), or fitted affinity estimates that change substantially with the concentration range, may indicate model misspecification, non-equilibrium behavior, heterogeneity, or cooperative binding. Differences between assay conditions, uncertain free concentrations, and uncertainty in fitted \(K_d\) values can also change the predicted selectivity. In such cases, compare like with like, examine uncertainty, and use a binding model appropriate to the evidence rather than treating the ratio as an exact property of every setting.

The affinity ratio is a compact way to rank binding strength, while matched-concentration occupancy is more directly interpretable for a particular exposure. Occupancy difference is useful when the practical concern is separation in percentage points; the occupancy-odds ratio is useful when relating the comparison directly to the affinity ratio under the one-site model. These summaries are not interchangeable. A plot of each target’s occupancy against log free concentration can reveal where both are unsaturated, where their occupancy difference is largest, and where both approach saturation.

The independent-target equation also does not account for other ligands competing for the same binding site. If competition or ligand depletion is important, \(C/(K_d+C)\) may not represent occupancy, and the simple cross-target comparison can fail. That is a reason to use a model that includes the relevant binding processes, not a reason to interpret a simple-model affinity ratio as an observed occupancy ratio. Nor does fractional occupancy by itself establish a functional response: biological effects can depend on target abundance, signaling, and other mechanisms.

Common Pitfall

Incorrect reasoning: “Target A has tenfold higher affinity, so it will have tenfold higher fractional occupancy at every concentration.” This confuses the affinity ratio with the fractional-occupancy ratio. Under the stated model, the ratio of occupancy odds equals the affinity ratio; fractional occupancies themselves change nonlinearly with concentration and both approach one at high concentration.

Correction: Specify the matched free concentration, calculate each occupancy separately, and report the comparison that answers the question. State the direction of the \(K_d\) ratio, and do not infer an occupancy difference from the affinity ratio alone.