Integrating Binding and Target-Engagement Assays

Why These Assays Need a Shared Model

Binding affinity, receptor occupancy, and downstream target engagement are related, but they are not the same quantity. A binding assay informs how strongly a ligand associates with a target under specified conditions. An occupancy assay aims to quantify the fraction of target bound in a biological system. A downstream assay measures a response that may depend on occupancy, but can also depend on signaling, feedback, and other biology. Combining these sources requires an explicit model linking them rather than treating their reported values as interchangeable.

The practical task is to identify, for each assay, both the biological quantity it informs and the observation process that turns that quantity into a recorded value. The calibration can then use shared biological parameters where justified, while retaining assay-specific calibration and error terms. This is consistent with the earlier material on linking measurements to latent model states and modeling assay-specific observation error.

From Binding to Fractional Occupancy

This relationship assumes one-to-one binding, a single affinity, equilibrium, and a known free ligand concentration. It also assumes that the ligand and target conditions in the model correspond to those in the assay. In particular, using total ligand concentration in place of free concentration is justified only when ligand depletion is negligible or when the model explicitly accounts for depletion.

The formula follows by expressing the total target concentration as \(R_{\mathrm{tot}}=[R]+[CR]\). Since \(K_d=[C][R]/[CR]\), rearrangement gives \([CR]/[R]=C_{\mathrm{free}}/K_d\). The occupied fraction is therefore:

Thus \(K_d\) is the free concentration at which half the target is occupied in this model: substituting \(C_{\mathrm{free}}=K_d\) gives \(\theta=1/2\). This interpretation is conditional on the equilibrium and single-site assumptions; it is not a universal conversion from a reported assay concentration to in vivo occupancy.

If ligand depletion is material, free and total concentrations differ. For a closed system with total ligand \(C_{\mathrm{tot}}\), total target \(R_{\mathrm{tot}}\), and complex concentration \(x=[CR]\), mass balance gives \(C_{\mathrm{free}}=C_{\mathrm{tot}}-x\) and \(R_{\mathrm{free}}=R_{\mathrm{tot}}-x\). Substitution into the definition of \(K_d\) yields:

The physically meaningful solution is the smaller root, which lies between zero and \(\min(C_{\mathrm{tot}},R_{\mathrm{tot}})\). In that setting, occupancy is \(x/R_{\mathrm{tot}}\), not the expression obtained by substituting total ligand concentration into the negligible-depletion formula.

Map Each Assay to Its Observation Process

A binding assay commonly records a signal related to bound target over a concentration series. After any justified background or nonspecific-binding correction, a simple observation model is \(Y_{\mathrm{bind}}=B_{\max}\theta+\epsilon_{\mathrm{bind}}\), where \(\epsilon_{\mathrm{bind}}\) represents assay error. If the recorded signal has a baseline or scale factor, those features belong in the observation model rather than being silently attributed to affinity. The assay may identify \(K_d\) only to the extent that the concentration range and measurements distinguish the curve's shape from its capacity and background.

An occupancy assay aims to measure \(\theta\), but its recorded value may be a transformed signal. For example, a calibrated assay might satisfy \(Y_{\mathrm{occ}}=a+b\theta+\epsilon_{\mathrm{occ}}\), where \(a\) is a baseline and \(b\) is a scale. A direct report of “percent occupied” is already an assay-derived estimate; it is not automatically the model's error-free occupancy. Calibration uncertainty, sample handling, and specificity can affect the mapping.

A downstream target-engagement readout requires another biological link. For a simple inhibitory response, one possible model is \(M=M_0-E_{\max}\theta\), where \(M_0\) is the baseline measurement and \(E_{\max}\) is the maximum reduction in the measured response. This is an example, not a universal law: the response may be nonlinear in occupancy, may have a nonzero residual level, or may reflect signaling steps that are not measured by the occupancy assay. The downstream instrument then observes \(M\) through its own scale, baseline, and error process.

Consequently, a joint calibration can share \(K_d\) through the occupancy equation while assigning each source its own observation model. A binding measurement informs the affinity and binding capacity; an occupancy measurement informs the occupied fraction, subject to its calibration; and a downstream measurement informs the biological mapping from occupancy to the measured response. A single assay should not be counted twice merely because its processed value is also used as an input to another reported quantity.

Linking Occupancy to a Downstream Readout

A downstream measurement can be informative about target engagement, but only through a specified biological relationship. For example, a decrease in a phosphorylation signal might be consistent with greater occupancy, yet the signal could also depend on pathway amplification, receptor abundance, or compensatory processes. The model must state which mechanisms are represented and which are absorbed into empirical parameters. The proportional model below is useful as a transparent starting point when its assumptions are plausible.

Assumptions, Diagnostics, and Alternatives

The shared model is most defensible when binding and occupancy refer to the same molecular target and ligand, units and concentration definitions are harmonized, and the conditions governing affinity are sufficiently comparable. Equilibrium occupancy additionally requires that the system has had time to equilibrate and that the relevant free concentration is represented. The downstream link requires that the measured response is meaningfully related to the modeled occupancy under the experimental conditions. These are scientific assumptions to evaluate, not consequences of fitting the equations.

Useful diagnostics follow the model's structure. Examine binding residuals across concentration: systematic curvature, an apparent nonzero lower asymptote, or a poor fit near saturation may indicate background, multiple binding components, or an unsuitable one-site model. Compare occupancy predictions and observations across concentrations or conditions: consistent offsets can point to assay calibration, while concentration-dependent disagreement may suggest an incorrect affinity or free-concentration assumption. For downstream data, residual trends with occupancy or baseline may indicate that a linear response mapping is inadequate. If all sources appear to disagree in the same direction, first check shared inputs and biological assumptions; if only one source disagrees, inspect its observation process and sample handling.

The simpler separate or empirical approaches can be preferable when the data do not support the mechanistic links, when assay conditions are not comparable, or when important components are unmeasured. A fitted joint model does not prove that its mechanism is correct. Its value is that it makes the proposed links explicit, testable, and available for calibration and prediction.