Glide seems to fail to produce accurate docking poses when flexible ligands (e.g. 22 rotatable bonds) are treated. What can I do to get better poses?

Docking such flexible ligands is extremely challenging, due to the very large configuration space. First, try keeping multiple poses per ligand, so you can look at the other good poses that Glide is finding. Perhaps Glide is producing a native-like pose, but it is not scoring as well (i.e., worse Emodel, if using SP) as the top pose. If there are native-like poses, then this is a scoring issue, rather than a sampling issue.

If the conformation generator is producing native-like ligand conformations, then it's possible that these are being eliminated at the intermediate docking stages. Increasingly the docking funnel width could help in this case. Under "Advanced Settings" on the Settings tab, there are some parameters governing the number of intermediate poses kept, and increasing these might help.

If the problem is sampling, you could generate conformations with MacroModel or ConfGen and then dock them rigidly with Glide. You could also use torsional constraints (from Suite 2010 on) to limit the number of torsional degrees of freedom, if you know that some torsions should adopt a given conformation.

Keywords: sampling, flexible ligands, Glide, docking

Back to Search Results