Molecular Glue-Induced Protein Interactions: Assay Design and Specificity Controls

A compound-dependent recruitment signal is not, by itself, proof of a molecular glue mechanism. The signal may reflect a specific interaction, a change in reporter geometry, nonspecific protein association, or interference with the detection chemistry. Conversely, an inactive biochemical assay may lack the protein domain, accessory subunit, or concentration range required to reproduce recruitment observed in cells.

Molecular glue characterization therefore begins with the protein pair and the conclusion required from the experiment. For early research programs, the central tasks are to establish compound-dependent complex formation, separate binding affinity from interaction stabilization, test recruitment specificity, and determine which additional measurements are needed to explain a cellular phenotype.

Molecular Glue-Induced Protein Recruitment

Molecular glues induce or stabilize protein–protein interactions. The recruited partner may be an E3 ligase component, but degradation is not the defining property of every molecular glue. Stabilization of an existing interaction can also alter sequestration, activity, or signaling. A characterization plan should name the proteins and intended functional consequence rather than assume every system follows the same target–degrader–E3 pathway [1,2].

Three measurements should remain distinct. Binary binding describes association of the compound with an individual protein. Recruitment describes a compound-dependent increase in association between two proteins. Cooperativity describes the thermodynamic coupling between the relevant binding events under a specified model. A compound can bind strongly to one protein but recruit the second poorly; a weak binary binder can become substantially stabilized in the assembled complex.

The baseline protein–protein interaction is equally important. If the proteins already associate without compound, the assay measures enhancement of that interaction, not necessarily creation of a new one. A proximity readout may contain contributions from both the basal protein complex and the compound-containing complex. If the baseline interaction is not detectable, it should be reported as below the method's measurement range rather than assigned an exact dissociation constant of infinity [1].

Unlike a bifunctional PROTAC, a molecular glue does not require two independently optimized binding groups connected by a linker. The binding routes and accessible equilibria may therefore differ. A PROTAC model that assumes appreciable compound binding to both isolated proteins should not be transferred unchanged to a glue that binds predominantly to one partner or to a composite interface [1,2].

These distinctions determine the first experiment. If a project already has robust cellular degradation but no defined partner, target or effector identification may precede purified-protein recruitment studies. If the proteins are known but a plate assay is ambiguous, a focused counterscreen and an orthogonal measurement are usually more informative than immediately collecting a broad panel of structural techniques. A molecular glue interaction characterization study can be scoped around that specific uncertainty.

Protein Constructs and E3 Ligase Complex Assembly

Target Domains, Accessory Subunits, and Native Protein Partners

Construct selection affects whether the recruitment interface exists in the assay. A convenient soluble domain may omit a relevant helix, degron, flexible segment, or partner-dependent conformation. An isolated substrate receptor may also differ from its assembled E3 complex in folding, stability, accessibility, and substrate recognition. The minimum recruitment system should be established experimentally rather than equated with the smallest proteins available.

The DCAF15–RBM39 system illustrates why this matters. Structural and kinetic work examined E7820-dependent recruitment using DDA1–DDB1–DCAF15 and the RBM39 RRM2 domain, identifying a compound-modified interface that recognizes the substrate helix [3]. That study supports a defined assembly for that interaction. It does not imply that every E3-dependent glue requires those accessory proteins or that the recruitment assembly alone reproduces the full ubiquitination reaction.

Other glues require a target-associated partner. Cyclin K degraders studied by Kozicka and colleagues recruit DDB1 through a compound-bound CDK12–cyclin K complex [4]. Testing cyclin K alone would omit the compound-binding interface central to that mechanism. The protein whose abundance decreases in cells is therefore not always the isolated protein to use for the initial binary-binding experiment.

Record the sequence boundaries, tags, mutations, cofactors, and subunit composition for each preparation. Assess purity, soluble recovery, expected assembly, and a relevant activity or binding property. Oligomeric heterogeneity can alter apparent stoichiometry and avidity. A concentration series is useful when the proteins self-associate or the assembled receptor complex dissociates during dilution.

Tags, Immobilization, and Protein Competence

Tags and labels can occlude the recruitment surface, change oligomerization, or bring detection reagents together independently of the intended interaction. Where feasible, compare an alternative tag position, an untagged competitor, or a different capture configuration. Agreement between configurations is particularly valuable when the initial signal depends strongly on one tag orientation.

An interface mutant is a useful specificity control only when its integrity is established. Loss of recruitment could reflect disruption of the contact surface, but also misfolding, altered subunit assembly, or loss of compound binding. Check the property the mutation is intended to preserve: for example, receptor assembly or binary ligand binding. A mutant that fails these checks cannot isolate the recruitment interface as the sole cause of the result.

For a biochemical recruitment assay, the required components may stop at the substrate-recognition assembly. A subsequent ubiquitination assay additionally requires the appropriate E1, E2, ubiquitin, ATP, and catalytically competent E3 system. Relevant E3 activation states should be controlled when they affect that reaction. Recruitment and catalytic reconstitution answer different questions and should not be judged by the same positive-control criterion.

Biochemical Assays for Molecular Glue-Induced Complexes

TR-FRET and Fluorescence Polarization

TR-FRET is useful for screening compound-dependent proximity between labeled proteins or tag-binding reagents. FP can monitor binding of a labeled peptide to a larger partner. Both permit concentration-dependent comparisons with relatively modest sample consumption, but the measured signal depends on the reporter configuration as well as complex abundance [1,2].

For TR-FRET, inspect the donor and acceptor channels as well as the calculated ratio. A ratio change can arise from altered donor emission, acceptor interference, or a change in fluorophore geometry. The same complex can produce different maximal signals in different configurations. Normalization does not automatically make those signals proportional to the fraction of proteins in a complex.

For FP, increased anisotropy can reflect specific binding, but aggregation, changes in fluorescence intensity, or viscosity can also contribute. Retain total fluorescence as a diagnostic readout. A compound that alters the labeled peptide signal in the absence of the protein partner needs further investigation before its concentration response is treated as recruitment.

A useful initial design measures basal protein association, compound response, single-partner controls, and a competent reference system on the same preparation batch. Assay signal and reagent stability should be checked across the incubation window. Incubation order can be tested when compound binding is slow or one component is unstable, but order-dependent signals should not be described as equilibrium constants without establishing equilibration.

SPR for Recruitment and Dissociation Measurements

SPR can compare binding in the absence and presence of a molecular glue, provided the immobilized protein remains competent and the solution components are correctly matched. It can also examine association and dissociation of a recruited partner at defined compound concentrations. The configuration must distinguish binding of the compound itself from additional binding of the second protein.

A recruitment experiment may use one captured protein and a second protein titrated in solution, with controlled compound concentrations. Include a suitable reference surface, solvent correction where needed, compound-only injections, and baseline controls for the second protein. Check whether the response is compatible with the expected assembly rather than interpreting every positive response as the intended complex.

Dissociation depends on the running-buffer composition. Maintaining compound in the dissociation buffer measures behavior under continuing ligand exposure; removing it changes the equilibria and may combine ligand loss with partner dissociation. Rebinding or surface heterogeneity can further alter the observed decay. State the configuration before comparing off-rates or residence times between compounds. SPR interaction analysis is particularly relevant when the unresolved question is affinity stabilization or kinetic behavior rather than complex mass alone.

Native MS and Mass Photometry

Native MS can resolve protein assemblies and, when mass resolution permits, compound-associated states. Its value in a molecular glue study is composition: does a compound-dependent species contain the expected proteins, or is the signal more consistent with self-association, an alternative subunit combination, or nonspecific adduction?

Buffer exchange and ion transfer can change the population. Different complexes may ionize, transmit, or survive activation differently. A complex peak provides compositional evidence, but its ion fraction is not automatically a solution mole fraction or an affinity constant. Compound concentration controls, component-only spectra, expected masses, and source-condition checks strengthen the interpretation. Native ESI-MS for noncovalent complexes can therefore complement recruitment assays without being imposed as the first method for every glue series.

Mass photometry can examine protein assembly at suitable concentrations, but it generally cannot resolve the small mass increment of one molecular glue. Closely spaced protein assemblies can also overlap. A new higher-mass population may represent the intended heterocomplex, a homooligomer, or aggregation. Component controls and theoretical competing masses are needed, and measurement dilution must not be assumed to preserve the incubation equilibrium.

Orthogonal measurements should remove a specific ambiguity. Repeating the same tag-dependent proximity assay with a different emission setting is not as independent as combining it with an appropriately controlled composition or binding measurement. Conversely, two techniques that require different buffers may disagree because they interrogate different equilibria rather than because one assay is incorrect.

Molecular Glue Concentration–Response and Cooperativity

Compound Titration and Protein Concentration Matrices

The EC₅₀ of a recruitment curve depends on the protein concentrations, basal interaction, compound affinity, interaction stabilization, and assay readout. Two compounds tested at one protein concentration can have similar EC₅₀ values while differing in these underlying properties. Changing the protein concentration can shift the apparent ranking without changing the compounds themselves [1,2].

A dual titration varies the glue and one protein partner while holding the other at a defined concentration. The range should include informative basal association, compound-dependent enhancement, and a measurable plateau where feasible. An initial feasibility series should establish usable protein concentrations before a large compound matrix is committed. Concentrations copied from an unrelated target pair may produce either a saturated assay with little dynamic range or a signal too weak to fit.

If proteins substantially bind in the vehicle condition, the remaining assay dynamic range may be small. Reducing one partner can reveal stabilization, provided sensitivity and protein stability remain adequate. If binding is exceptionally tight relative to reagent concentration, apparent potency can become limited by stoichiometry and ligand depletion. Simply lowering protein concentration is not always possible because detection sensitivity or protein instability may set a practical lower limit.

Figure 1. A concentration matrix separates compound-dependent recruitment from effects of protein concentration and basal association. The schematic includes a vehicle column and a protein-partner titration at fixed concentration of the other partner. Its qualitative signal pattern is illustrative, not experimental data or a universal molecular glue response.

The 2025 work of Guilbert and colleagues examined how assay conditions and concentration-response span relate to glue-induced affinity shifts in a defined model system [2]. A separate study developed matrix-based models incorporating basal protein affinity, ligand affinity, and cooperativity, with validation using TR-FRET and FP [1]. These studies support mechanistic analysis when model assumptions are satisfied; they do not imply that any concentration curve uniquely determines all thermodynamic parameters.

Cooperativity Definitions and Model Identifiability

Define the convention before reporting α. In the 2025 molecular glue models, α can be expressed as the dissociation constant for protein association with bound glue divided by the basal dissociation constant. Under that convention, α below 1 indicates stabilization, and the affinity improvement is 1/α. Other literature uses an enhancement factor with the opposite direction [1,2]. A numerical α value without its defining ratio is not suitable for cross-study comparison.

State which protein binds the glue, which basal interactions are included, and whether the measured signal reflects the compound-containing complex alone or both basal and stabilized complexes. The fitted model should respect reagent mass balance. Under tight-binding conditions, free and total concentrations cannot be treated as interchangeable; using a simple independent binding curve can generate misleading parameters.

Fit quality is necessary but not sufficient. Distinct parameter combinations can produce similar curves if the matrix lacks informative concentrations or if a transition is not sampled. Inspect parameter uncertainty, sensitivity to the selected concentration window, and agreement across independent preparations. An unmeasurable basal interaction may sometimes be constrained through joint modeling, but a precise value should not be reported when the data permit only a broad range.

A compound response that has not reached equilibrium needs a kinetic interpretation. Slow-binding reversible or covalent mechanisms can produce incubation-time-dependent potency. The equilibrium models discussed here should not be applied unchanged to an irreversible reaction. Record the preincubation sequence and demonstrate whether the selected time window gives stable behavior before using equilibrium terminology [1].

Interpreting Plateau Changes and High-Concentration Effects

The curve plateau is not a universal measure of glue quality. It can vary with tag geometry, active protein fraction, detector saturation, background subtraction, and the concentration of the limiting partner. Compare compounds within a qualified configuration and investigate marked changes in the raw channels before interpreting a different plateau as greater complex formation.

Many monovalent glue models do not predict the classical dual-warhead PROTAC hook effect. Nevertheless, a declining assay signal at high compound concentrations can occur through solubility problems, detector interference, aggregation, competing interactions, or a mechanism not represented in the selected model. Neither the presence nor absence of a downturn alone establishes molecular glue activity [1,2].

For an analogue series, retain both fitted potency and the measurement conditions. Where warranted, separate ligand affinity from the induced protein-affinity shift. If those parameters cannot be resolved, report an assay-specific recruitment ranking and its uncertainty rather than relabeling EC₅₀ as an intrinsic affinity or cooperativity value.

Selectivity Controls and Assay Interference

Component, Compound, and Interface Controls

Vehicle controls establish basal association under the same solvent conditions as the compound wells. Single-partner controls assess whether the compound affects a protein, tag, or detector without the intended complex. A competent reference compound or reference interaction helps distinguish a failed reagent preparation from an inactive candidate. Interpretation depends on the combined control results rather than any single comparison.

An inactive analogue is especially useful when it preserves relevant physicochemical behavior while disrupting the molecular feature needed for recruitment. Differences in solubility, permeability, or fluorescence can otherwise confound the comparison. If the analogue also loses binary binding, it tests a broader requirement than the protein–protein interface alone. Where possible, pair it with a protein-interface perturbation that retains the required protein competence.

Unlabeled competition can help evaluate a tagged recruitment system. However, competition may displace the compound-binding partner, alter subunit assembly, or change free protein concentrations. Its interpretation should follow the proposed mechanism rather than assume every reduced signal is evidence of a unique binding site.

Figure 2. Specificity assessment combines component dependence, a competent interface control, and an orthogonal readout. Reporter interference and nonspecific association are alternative explanations to test, not outcomes inferred from one negative control. The diagram represents experimental comparisons rather than measured performance.

Fluorescence, Aggregation, and Compound Solubility

Inspect compound-only and detector-only behavior over the same concentration range used for recruitment. Absorbance, quenching, autofluorescence, scattering, and disruption of tag-binding reagents can create concentration-dependent artifacts. Time-resolved detection reduces some background contributions, but does not make TR-FRET immune to compound interference. Non-cellular PPI assay guidance emphasizes controls for the assay format and detection reagents [7].

Aggregation can nonspecifically associate proteins and increase apparent proximity. Concentration-dependent turbidity, broad higher-mass species, unusual dependence on protein loading, and sensitivity to handling conditions are warning signs. Solubility should be examined in the final assay buffer, not inferred from a clear concentrated DMSO stock. Keep final solvent content matched across the concentration series and assess material recovery when precipitation is suspected.

A compatible detergent or altered salt condition can help investigate nonspecific association, but the same perturbation can disrupt a genuine interaction. Loss of signal after detergent addition is therefore not a definitive aggregation diagnosis. Combine the result with component dependence, soluble protein recovery, and an independent measurement that addresses the proposed failure mechanism.

Neosubstrate Selectivity and Interface Confirmation

Selectivity should be evaluated against meaningful competitors. A recruitment series aimed at one neosubstrate should include relevant homologous domains or known substrates where material is available, not just an unrelated protein that cannot form the proposed interface. A negative result for a related substrate is informative only if that substrate is competent and the assay can detect its recruitment when appropriate.

The discovery of the IKZF2-selective glue NVP-DKY709 illustrates recruitment-guided selectivity evaluation against related IKZF proteins, supported by structural analysis of the compound-induced complex [6]. The important lesson is to connect selectivity to the interacting protein features and supporting measurements. Its compound-specific selectivity cannot be generalized to all CRBN binders or all assay constructs.

A 2026 affinity-selection MS study compared ligand enrichment with and without protein partners and followed selected compounds with orthogonal proximity and cellular assays [5]. Its enrichment thresholds belong to that workflow. Affinity-selection MS measures retained ligand after selection; it is not the same measurement as native MS of intact complexes. The two should not be conflated simply because both use mass spectrometry.

Where an interface remains uncertain, structural analysis, selective mutagenesis, and HDX-MS conformation and dynamics measurements can provide complementary information. Reduced exchange can reflect direct interface burial or a more distant conformational change. Protection should be evaluated alongside peptide coverage, complex occupancy, and appropriate binary-component comparisons before it is assigned as a direct contact.

Protein Recruitment, Ubiquitination, and Cellular Degradation

Recruitment and Catalytic Productivity

For a degrader, recruitment is one step in a longer mechanism. The assembled complex must permit ubiquitin transfer and subsequent handling by the cellular degradation machinery. Productive positioning, accessible substrate residues, E3 catalytic competence, and cellular availability can separate a strong recruitment signal from a useful degradation response.

A reconstituted ubiquitination assay can test this transition using competent E1, E2, E3, ubiquitin, ATP, and substrate components. Controls should establish dependence on the compound and necessary reaction components. Substrate-associated ubiquitination must be distinguished from modification of the E3 itself or another reaction component. A change in total reaction signal is not automatically evidence that the intended target was modified.

If a substrate-binding domain supports recruitment but lacks the relevant ubiquitination context, use it to answer the binding question and assess a suitable full-length or larger construct for the catalytic step. Conversely, a catalytic failure should not invalidate a well-supported recruitment result when the enzyme system is incomplete. Specify which step each experimental configuration can reproduce.

Cellular Dependence and Conflicting Readouts

Cellular target reduction can arise through direct degradation, reduced synthesis, altered transcription, or secondary stress responses. Measure endogenous target abundance where feasible and include cell number or viability controls. A tagged reporter can differ from the endogenous protein in localization, accessibility, turnover, and expression level; reporter loss alone needs appropriate confirmation.

Genetic perturbation of the proposed effector, a suitable rescue experiment, and a substrate variant can help establish pathway dependence. Proteasome inhibition can support a proteasome-dependent process, but it also changes cellular protein homeostasis. An inhibitor rescue without recruitment or pathway-specific evidence does not uniquely identify the compound's direct mechanism. For non-degrading glues, replace the degradation endpoint with the relevant activity, sequestration, or signaling measurement.

For compound optimization, compare biochemical recruitment and cellular results under explicitly defined conditions. DC₅₀, maximum target reduction, incubation time, and target recovery after washout describe different aspects of cellular behavior. A slow dissociation rate measured on an immobilized biochemical system is not a universal predictor of those endpoints. Cyclin K studies that combined recruitment, structural, and cellular measurements demonstrate the value of evaluating these steps together [4].

The most informative interpretation identifies where the evidence is concordant and where it stops: specific compound-dependent recruitment, a qualified affinity shift, an assigned protein assembly, substrate ubiquitination, or pathway-dependent target reduction. If the biochemical data support only assay-specific recruitment potency, retain that description rather than assigning an unsupported Kᴅ or cooperativity factor.

For an early molecular glue series, a practical progression is to establish competent protein components and controlled recruitment, resolve the parameter or specificity question that separates the analogues, and then test the relevant cellular mechanism. Additional methods are justified when they explain a remaining discrepancy. This keeps the analysis connected to compound selection and mechanism research without turning a positive proximity assay into a broader biological claim than the data support.