PROTACs vs molecular glues: what docking can tell you

PROTACs and molecular glues recruit E3 ligases to degrade target proteins instead of merely inhibiting them. Both modalities have entered clinical trials, but their mechanisms differ sharply—and so do the computational questions you can answer with conventional docking versus those that require ternary-complex modeling.

Why degraders are not just bigger inhibitors

A PROTAC (proteolysis-targeting chimera) is a heterobifunctional molecule: one end binds your target protein, the other end recruits an E3 ubiquitin ligase (most commonly cereblon or VHL), and a flexible linker spans the two. When both binding events occur simultaneously, the target is ubiquitinated and sent to the proteasome. The pharmacology depends on catalytic turnover—one PROTAC molecule can degrade many copies of the target—so low occupancy can yield high efficacy.

Molecular glues also induce proximity between a target and an E3 ligase, but they do so by stabilizing a protein–protein interface that would otherwise be weak or absent. They are typically smaller, monovalent molecules that bind at or near a surface patch, creating a neo-interface. Lenalidomide and pomalidomide are well-known examples that glue cereblon to Ikaros-family zinc fingers. Unlike PROTACs, molecular glues do not have two discrete pharmacophores joined by a linker; the entire molecule participates in the ternary interaction.

What binary docking can tell you: the warhead

Classic two-body docking—ligand into a target binding site—remains valuable for the target-binding moiety (the warhead) of a PROTAC. If you are starting from a known kinase inhibitor or bromodomain ligand, you want to confirm that attaching a linker at a given position does not disrupt key hydrogen bonds or bury solvent-exposed vectors. Liganx Studio lets you dock candidate warheads rapidly and inspect which attachment points leave the binding mode intact.

For molecular glues, binary docking into the E3 ligase (or occasionally the target) can identify whether your candidate occupies the expected pocket—cereblon's β-hairpin cage for IMiD derivatives, for example—but it cannot predict whether the compound will create a productive neo-surface. That requires modeling the ternary state.

The ternary complex problem

Successful degradation depends on geometry: the E3's catalytic machinery must be oriented correctly relative to accessible lysines on the target. A PROTAC with excellent binary affinity for both endpoints can fail if the linker length or topology forces the two proteins into an unproductive arrangement. This is why structure-activity relationships for degradation often show sharp cliffs with linker length—adding or removing a single ethylene glycol unit can abolish activity.

Ternary-complex modeling approaches range from template-based (when a crystal structure of a related PROTAC ternary complex exists) to de novo methods like PROTAC-TS or PRosettaC that sample linker conformations and protein orientations. AlphaFold-Multimer has also been applied, though its accuracy for induced interfaces mediated by a small molecule is still being benchmarked. The key output is not just a pose but a cooperativity metric: does the PROTAC stabilize the ternary complex relative to the sum of the binary interactions?

Cooperativity and the hook effect

Positive cooperativity—where ternary Kd is tighter than predicted from binary affinities—is the hallmark of a good PROTAC. Conversely, if the PROTAC binds target and E3 independently but the ternary complex is strained, you may see a hook effect: degradation peaks at low-nanomolar concentrations then diminishes at higher doses as binary complexes (PROTAC:target or PROTAC:E3) outcompete the productive ternary species.

Binary docking workflows cannot detect cooperativity because they evaluate each binding event in isolation. If you observe non-monotonic dose-response curves in degradation assays, consider whether your linker geometry permits a stable ternary arrangement or whether steric clashes force the complex apart.

When you still need binary docking for degraders

Even in PROTAC campaigns, binary docking has three practical uses. First, when selecting or modifying the warhead—especially if you are docking against mutant targets to anticipate resistance—you need to ensure the target-binding mode is preserved. Second, when choosing the E3 ligand (cereblon vs VHL vs others), docking the ligase-binding moiety confirms you have not broken an essential interaction. Third, binary docking can rule out obvious off-targets by screening the PROTAC or molecular glue against a panel of unintended proteins.

For molecular glues, binary docking is even more limited: you can verify that the compound fits into the cereblon or other E3 pocket, but predicting which target proteins will be recruited requires either structural templates (from cryo-EM or crystallography of ternary complexes) or proteomics-driven discovery followed by retrospective structural validation.

Resistance mutations in degrader modalities

Resistance to PROTACs can arise in the target (mutations that abolish warhead binding), in the E3 ligase (cereblon mutations are well-documented in multiple myeloma after IMiD exposure), or via upregulation of deubiquitinases. Docking the warhead against common target resistance mutations—using the same workflow you would for a classical inhibitor—helps anticipate whether the PROTAC will retain binary affinity. However, a mutation that preserves binary Kd but alters surface geometry may still block ternary-complex formation and abrogate degradation.

Molecular glues face analogous risks: mutations at the neo-interface (either on the target or the E3) can disrupt the induced protein–protein contact without necessarily affecting the binary affinity of the glue for cereblon or VHL. This makes rational design of next-generation glues particularly challenging.

Practical workflow: combine binary and ternary methods

A pragmatic computational strategy starts with binary docking to validate or optimize each pharmacophore, then escalates to ternary modeling once you have a candidate linker. Use binary docking in Liganx to iterate on warhead substitutions, confirm exit vectors for linker attachment, and screen for off-target liabilities. When you have a shortlist of linker lengths and chemistries, apply a ternary-complex prediction tool to rank candidates by cooperativity and to identify which lysines are proximal to the E3 catalytic site.

For molecular glues, the workflow is inverted: structural biology (cryo-EM or X-ray of ternary complexes) usually comes first, followed by binary docking to explore analogs that preserve the key neo-interface contacts. Computational prediction of entirely new glue targets remains an open problem, though machine-learning models trained on known cereblon and VHL substrates are beginning to show promise.

What to expect in the next five years

Ternary-complex prediction is improving rapidly. AlphaFold3 and related models that explicitly handle small-molecule cofactors may eventually predict PROTAC ternary structures de novo with acceptable accuracy. Cooperativity scoring functions that account for both enthalpic and entropic contributions are still under development; most current tools report a geometric feasibility score rather than a true delta-delta-G.

For molecular glues, the bottleneck is target discovery: once you have a ternary structure, analog design follows established medicinal-chemistry principles. High-throughput structural proteomics—using cryo-EM to screen which endogenous proteins co-purify with a glue-ligase complex—may accelerate the identification of new substrates and expand the degrader toolkit beyond the cereblon-IMiD paradigm.

Conclusion

Targeted protein degradation is mechanistically distinct from inhibition, and the computational tools must match the biology. Binary docking remains essential for warhead optimization and off-target profiling, but it cannot predict cooperativity, ternary geometry, or degradation efficiency. Combine binary docking in Liganx Studio with ternary-complex modeling to build a complete picture—and remember that even the best computational model must be validated in a cellular degradation assay, because ubiquitination efficiency and proteasomal processivity lie beyond the reach of docking scores.

References

• Gadd MS, Testa A, Lucas X, et al. Structural basis of PROTAC cooperative recognition for selective protein degradation.Nat Chem Biol. 2017;13(5):514–521. doi:10.1038/nchembio.2329
• Békés M, Langley DR, Crews CM. PROTAC targeted protein degraders: the past is prologue.Nat Rev Drug Discov. 2022;21(3):181–200. doi:10.1038/s41573-021-00371-6
• Chamberlain PP, Hamann LG. Development of targeted protein degradation therapeutics.Nat Chem Biol. 2019;15(10):937–944. doi:10.1038/s41589-019-0362-y
• Zeng M, Xiong Y, Safaee N, et al. Exploring targeted degradation strategy for oncogenic KRASG12C.Cell Chem Biol. 2020;27(1):19–31.e6. doi:10.1016/j.chembiol.2019.12.006
• Bai L, Zhou H, Xu R, et al. A potent and selective small-molecule degrader of STAT3 achieves complete tumor regression in vivo.Cancer Cell. 2019;36(5):498–511.e17. doi:10.1016/j.ccell.2019.10.002