Glide Constraints

A Glide constraint is a ligand-receptor interaction requirement. The constraint usually means that a ligand atom must lie within a certain region defined in relation to features of the receptor that are responsible for ligand binding. To use Glide constraints, you must specify receptor sites for possible ligand interactions when you set up a receptor grid generation job. When you run a ligand docking job, you can select Glide constraints to apply from the list of receptor constraint sites that you defined for the receptor.

Other constraints that reflect ligand-receptor interactions indirectly are applied during docking. These include superposition of the ligand core on a reference core, and restriction of specified ligand torsions.

In Glide constraint docking jobs, Glide incorporates satisfaction of these constraints into several of its hierarchical filters, allowing prompt rejection of docked poses that fail to meet the requirements.

The first constraint filter is a simple one involving the atoms of the ligand. If a ligand does not contain atoms of the right types to make the required interactions with the receptor constraint atoms, Glide simply skips that ligand—for instance if a selected receptor constraint atom is a polar hydrogen and the ligand has no hydrogen-bond acceptors. If there are qualifying atoms in the ligand, Glide keeps a list of the possible “partner atoms” for each constraint, for use in subsequent filters. For hydrophobic constraints, this filter checks that the ligand contains a sufficient number of hydrophobic heavy atoms; the sum of the minimum numbers specified for all the selected constraints of this type. In addition, Glide locates centroids of ligand hydrophobic groups, which take the role of partner atoms in subsequent filters.

Other filters operate further down the funnel. These include matching distances between partner atoms (or hydrophobic group centroids) for different constraints against the corresponding distances between receptor atoms; matching distances from the ligand center to partner atoms against those from receptor site points to constraint atoms, and restricting orientations of, and rotations about, the ligand diameter to those that bring partner atoms into appropriate proximity with receptor atoms or regions. In addition, a restraining potential is used in grid optimizations (for hydrogen-bond and metal constraints only) to bring or keep constraint-satisfying ligand-receptor atom pairs at appropriate distances, and torsional sampling moves are not tried if they would move any ligand atom currently in a constraint-satisfying position. Finally, grid-optimized poses are rejected if they don’t strictly satisfy all selected constraints.

Core constraints are applied by a “snap and refine” algorithm. The core pattern is identified in the ligand, and the matching atoms are placed at exactly the same coordinates as the reference ligand. The rest of the molecule is rebuilt, preserving the internal coordinates of the non-core atoms. The rotatable bonds are then sampled, and passed through rough scoring, but the refinement part is skipped, as the core is placed exactly and no translation of the ligand is needed. The refinement (grid minimization) is performed, in which the non-core torsions are minimized, and sampling of torsions is then applied to improve poses, if possible. Finally, post-docking minimization is performed, with steep, flat-bottomed constraints on the core atoms. This last step may change the RMSD with respect to the reference core; prior to that, it is zero.

For information on using Glide constraints, see the Receptor Grid Generation Panel and Ligand Docking Panel DEPRECATED help topics.