GOLD is able to perform automated docking in the following special cases:
GOLD models positively-charged metal ions like hydrogen bond donors and attempts to reproduce observed coordination geometries by fitting acceptors to coordination points, using mappings within the GA chromosome and least-squares techniques.
Currently, GOLD is able to attempt docking to five metal ions: Mg, Zn, Fe, Mn and Ca. Testing has verified GOLD's ability to dock ligands into binding sites containing such ions.
GOLD does not need any special instructions to dock to metal ions: they are handled automatically if they are present in the binding site. However, there are some restrictions which the user should be aware of. Importantly, the metal ion must be coordinated to at least two protein atoms, in a reasonable geometry, so that GOLD can make a good guess at the coordination geometry.
At present, GOLD can only deal with two coordination geometries: octahedral and tetrahedral. However, one site may be shared by two acceptors if they are bonded to the same atom (for example one site can be shared by the two oxygens of an acid residue). This scheme allows the reproduction of good geometries for zinc with coordination number of 5 and calcium in EF hands (coordination number 7). Nevertheless, the current scheme is clearly inadequate for reproducing many of the coordination patterns that have been observed in crystal structures in the Cambridge Structural Database.
In the input file to GOLD, the metal ion should not have any bonds to coordinating atoms: if these are present in the PDB file you will have to delete them.
The output file ga_protein.mol2 will contain a number of dummy atoms representing idealised coordination positions. These dummy atoms will be connected to the metal ion.
In order to model metal ions within SYBYL, you need to load a parameter file, otherwise all metal ions will be typed as dummy atoms. Add the following line to your ~/.sybylrc file:
parameter open $TA_ROOT/demo/metals.tpd |
Note that there may be problems in the way that SYBYL handles metal ions: they are not always well behaved in the minimiser, and typically have valencies of 4 or 6, which may mean that hydrogens are added to the metal when you add hydrogens to the protein.
GOLD is able to dock covalently bound inhibitors. However, the software does not elucidate covalent binding: the user must tell GOLD which ligand and protein atoms are covalently bonded.
It is assumed that there is one atom linking the ligand to the protein (e.g. the O in a serine residue). Both protein and ligand files are set up with the link atom included (so, if the serine O is the link atom, it will appear in both the protein and ligand input files). Since the link atom will have a free valence there may be an error message from the type-checker.
Inside the GOLD least-squares fitting routine, the link atom in the ligand will be forced to fit onto the link atom in the protein. In all other respects, GA H-bond mapping and fitting remains unchanged.
In order to make sure that the geometry of the bound ligand is correct, the angle-bending potential from the Tripos Force Field has been incorporated into the fitness function. On decoding the docked ligand, the angle-bending energy for the link atom is added to the internal energy of the ligand.
This seems to work well in the systems on which GOLD was validated. However, since, at present, the protein is held rigid (apart from hydroxyl hydrogens), it does require that the position of the link atom in the protein is well defined.
In order to set GOLD up to perform covalent docking, the ligand and protein files must first be set up as described above. Then:
|
| Bring up the front end and define the problem in the normal way; then, click on the Covalent button. |
The covalent atom input boxes will then appear. Enter the atom numbers (as defined in the SYBYL MOL2 input files, or use PDB sequence numbers if PDB input is used) of the link atom for the protein and ligand into the appropriate box.
GOLD allows docking to be performed under constraints. The following constraints are provided:
Distance Constraints
Distance constraints can be incorporated into the GA fitness function. Any distance between a ligand and protein atom (or between two ligand atoms) can be constrained to lie between maximum and minimum distance bounds. If (when decoding a chromosome) a distance is found to lie outside its constraint bounds, a spring energy term is used to reduce the fitness score:
where x is the difference between the distance and the closest constraint bound and k is a user-defined spring constant.
Hydrogen Bond Constraints
This constraint forces GOLD to form a hydrogen bond between a protein atom and a ligand atom. One atom should be a donor hydrogen and the other atom should be an acceptor. The protein atom should be available for ligand binding (e.g. solvent accessible).
The constraint is incorporated into the least-squares fitting routine used by GOLD. Thus, when least-squares fitting is used to dock the ligand (by attempting to form hydrogen bonds encoded within the chromosome) the constraint is added to the least-squares mapping. The constraint has a weight of 5 relative to a hydrogen bond taken from the chromosome.
| Bring up the front end and define the problem in the normal way; then, click on the Edit Constraints button. |
This brings up the Constraint Editor window. Clicking on the Add button in this window will give you a choice of adding a Distance Constraint or an H Bond Constraint.
Distance Constraints Form
Fill in the form to specify the distance constraint, using the atom numbers as defined in the SYBYL MOL2 input files. If PDB input is used, use the sequence number. Distances are in Angstrom.
H-Bond Constraints Form
Fill in the form to specify the hydrogen bond constraint, using the atom numbers as defined in the SYBYL MOL2 input files. If PDB input is used use the sequence number.
Back to the main GOLD page