Generic selectors
Exact matches only
Search in title
Search in content
Search in posts
Search in pages
Select Page


Restrain Ligands is an option that can be activated during Repack, Minimize, Loop Rebuild, or Design. This is intended to freeze the molecule in place in relation to the protein during modeling. Warning: Only turn this on if you want the ligand to stay in place even if the physics of the interaction is wrong.


The Restrain Ligands option can be found whenever you are creating a Repack, Minimize, Loop Rebuild, or Design action. Once you have chosen the structure for an action and clicked the action name, a tab is open in the center window with options for that action. In the example below, we are preparing to run Repack on the structure from PDB ID 2YXJ which has a ligand that Cyrus renamed 143000 in chain Z.

By clicking , the ligands listed below this will be clickable. More than one ligand will appear if there are multiple in your file. You can select 1 or more. Then click .

What Does Restrain ligands Do?

Restrain ligands can be used if you find that your small molecule is moving too much during modeling. There are two main things that can cause this:

  1. This most commonly occurs because the ligand does not bind to the pocket in a favorable way because it lacks the cognate shape and chemistry to bind there. Rosetta responds to this by pushing it out of its original binding mode in order to relieve the unfavorable interactions. Rosetta will move the ligand until it finds the least unfavorable position in the region.
  1. Bad input can also kick the molecule out of its pocket. This could happen if the atoms in the ligand’s input structure are too far from ideal to create good parameters. The molecule could also have bad parameters if the OpenEye software we use to define the ligand has difficulty with your ligand’s chemistry. If bad input is the cause of bad movement, better input should be found before resorting to Ligand Restraints.

Whatever the cause, you can trick Rosetta into keeping the ligand in its current position with Restrain Ligands. This adds a penalty to the structure’s score whenever the ligand moves too far from its initial position. So modeling can still occur with normal sampling of the ligand’s conformations and positions, but it will bias sampling towards its initial position. This is essentially a virtual tether between the ligand and nearby protein atoms.

More specifically, a 1 angstrom tether is created between the C alpha atom of any amino acid within 4 angstroms of a non-hydrogen atom in the small molecule.

This doesn’t mean that a ligand can’t move more than an angstrom away. It just means that a larger step needs to by extra-favorable. Meaning, the large step must be a significant improvement in the structure’s overall score in order to overcome the penalty we created in Restrain Ligands. This does not freeze the ligand’s internal bonds, so conformer sampling will also still occur.

Cyrus is interested in feedback on Restrain Ligands.  If you feel they are too flexible, or too inflexible, please let us know; we are interested in tuning this tool based on your feedback.