Hints and tips
usage: sublime [options] -spheres pdbfile -site pdbfile > filtered_spheres.cex
Sublime reads spheres ("-spheres") from a pdb format file (only the coordinates and the radius are read; the radius is stored in the temperature factor field), along with the macromolecule ("-site") which was used to calculate the spheres (e.g. with sphinx).
Sublime removes spheres with a LIGSITE burial score (0 = completely exposed, 7 = complete buried) below the threshold specified by "-minscore". Remaining closely overlapped spheres are merged and buried spheres are removed. Surviving spheres are clustered into connected groups (where sphere surfaces are no farther apart than the threshold specified by "-slop") and sorted by the aggregate LIGSITE score of each cluster to select the most likely binding site(s).
Sublime's first sphere cluster usually (94 out of 123 cases tested) contains the correct binding site. Macromolecule or ligand atoms may be optionally specified in separate pdb files to define proximity ("-select_dist" or "-reject_dist") to include ("-select") or reject ("-reject") sphere clusters. Individual spheres within clusters may be selected or rejected by specifying "-byspheres". Use "-select" and "-whichcluster 1" with a bound ligand, if available, to guarantee that the correct sphere cluster is chosen. If no bound ligand is available, the default value of "-clusters" will save the top 5 scoring clusters, which should include the correct binding site(s). Cluster #1 is saved as residue #1, cluster #2 as residue #2, etc. Select a specific cluster by rerunning sublime with "-whichcluster".
Sublime can be used iteratively to select the appropriate spheres to define a site. Spheres may be optionally written out to a pdb format file ("-spheres_out"). A script for viewing the spheres in RasMol (sublime.rasmol) will also be created. Concatenate the protein, ligand, and sphere files together: % cat protein.pdb ligand.pdb spheres.pdb > foo.pdb. In rasmol, "source sublime.rasmol" will display the spheres. Spheres are color-coded by their LIGSITE scores to make it easy to alter the "-minscore" threshold if necessary:
Proteins with very shallow grooves or binding sites (e.g. PDB file 1ela) produce only a small number of spheres, which may be insufficient to contain a docked ligand. This is usually not critical, since the spheres are invisible during the second stage of docking and the ligand is free to move within the binding site, but performance will improve if the spheres accurately match the volume required for the docked ligand(s). If one or more bound ligands are available, their atoms can be used to define additional spheres by using the -addspheres filename option. filename is a PDB format file (containing one or more bound ligands) which is used to generate additional spheres at the atom positions. If the atoms have temperature factors, those are used for the sphere radii, otherwise a default radius of 1.5 is used. Manually placed spheres could also be supplied with the -addsphere option, but this is very seldom necessary.
LIGSITE: M. Hendlich, F. Rippmann, and G. Barnickel, J. Mol. Graphics, 15, 359-363 (1997).