Skip to main content

​
Introduction

Molecular glues function by inducing or stabilizing protein-protein interactions, forming ternary complexes between a target protein, the glue molecule, and a binding partner (such as an E3 ligase or other effector protein). Boltz Lab can facilitate modeling of these ternary complexes within the Sandbox and Design Projects by allowing you to specify multiple protein sequences.
To explore Boltz performance on your target, begin by adding two protein sequences in the Sandbox, and a molecular glue probe. Following the same steps as below for setting constraints (if needed) provides a good way to explore this target class, before beginning a Design Project.
This guide uses PDB entry 5HXB as a reference example of a molecular glue ternary complex.

​
Creating a Molecular Glue Target

​
Step 1: Set Up Design Project

When you create a Target within a Design Project, add two protein sequences instead of one:
  1. Protein 1: Your target protein (e.g., degradation target)
  2. Protein 2: Binding partner (e.g., E3 ligase, effector protein)
The platform will predict the ternary structure with both proteins present.
When you add your glue probe molecule, the model should predict a ‘pocket’ at the interface between the two proteins.

​
Step 2: Define the Binding Site

There are two approaches to defining where the molecular glue binds:
For well-formed interface pockets like PDB 5HXB, the pocket prediction will automatically identify the binding site at the protein-protein interface.This works when:
  • The proteins have a clear interface cavity
  • The glue binding site is well-defined
  • Interface residues are pre-positioned for binding
Molecular glue binding at protein-protein interface

​
Guiding Ternary Complex Formation

​
Using Contact Constraints

If the predicted ternary complex does not match your expected orientation, use the Contact Constraint to guide the model. The Contact constraint defines specific residue-residue interactions between the two proteins, helping to position them correctly relative to each other. When to use Contact constraints:
  • Proteins are predicted in the wrong relative orientation
  • Known key interface contacts are not formed
  • The glue binding site is not properly formed between proteins
Contact constraints work by specifying pairs of residues that should be in close proximity, helping the model to place the two proteins into the correct ternary complex geometry.

​
Key Considerations

Choose appropriate protein constructs:
  • Use domains that are known to interact (avoid including unstructured regions)
  • Match experimental constructs when available
  • Consider including known co-factors or modifications
The quality of interface pocket prediction depends on:
  • Clarity of the protein-protein interface
  • Presence of pre-existing cavity at the interface
  • Degree of induced fit required
If predictions are poor, try adjusting which interface residues you select for pocket definition.
Ensure your input molecule represents the actual glue structure:
  • Include all relevant functional groups
  • Use appropriate protonation states
  • Consider stereochemistry if relevant
Validate your predictions by:
  • Comparing to known ternary complex structures (when available)
  • Checking that predicted binding poses position the glue at the interface
  • Verifying that key interface contacts are maintained
  • Confirming predicted binding affinity is reasonable