Molecular Docking

The docking functionality uses a hybrid scheme of the FlexX docking and a force field (amber+gaff) optimisation.
The same scheme as employed by FlexX was used to generate the placement of the base fragment.
So first distance tables are generated to store all the distances between the interaction points in the protein.

The protein has to be supplied in a pdb format with hydrogens present. Molden can be used to generate such a pdb file.
For now, the orientation of for example OH groups (Serine,Threonine,tyrosine) have to be predefined.

The user can define a sphere around a to be picked point, which encloses all the interaction points which are relevant for the docking.
These interaction point can be either H(ydrogen Bond)_Acc(eptor), H(ydrogen Bond)_Don(onor) or aromatic/ch(2,3).
The user selected amino acid residues are then assigned interaction points.

For each combination of interaction groups, distances are collected and stored in a distance table.
Next a ligand must be supplied in ambfor .xyz format (molden own force field format), also charges must have been assigned for proper force field scoring.
Next interaction points are assigned to the ligand. From these, interaction triples are generated.
This list of triples is searched against the distance tables to find matching protein interaction triples, these are called base fragments.

For the ligand, two to three interaction points are generate for each OH group, One for each rotamer.
Initially the ligand is divided up into fragments. Fragments are typically connected by single SP3-SP3 or SP2-SP3 bonds.
These fragments are stored internally by molden.
This fragmentation is used to generate different conformations of the ligand by rotating around the dihedral angles of the single bonds separating the fragments.
The bonds that are not strict single bonds (SP2-SP3) are kept in order to carry out a complete scan of the energy with respect to these dihedrals.
For regular SP3-SP3 bonds the standard angles of the staggered conformations of ethane are employed.

When the debug button is pressed in the "Docking Start" window a file called "conformers.mol2" is written.
The orientation of the base fragment is very important for the following steps in the docking procedure.
That is why the base fragment placement is optimised to obtain a near perfect placement.

If we have a base fragment consisting of all Hydrogen bond acceptors and donors, the van der waals part of the force field scoring is reduced,
so that the orientation is dominated by electrostatic part of the force field scoring.

If we were dealing with a base fragment with one pair of Hydrogen bond acceptors or donors and one hydrophobic interaction point,
we rotate around the axis formed by this pair to find the energetically optimal angle for the remaining hydrophobic interaction point with respect to the pair.

Because the hydrophobic interaction between hydrophobic interaction points are much smaller than between a acceptor and donor pair
(round about seven times weaker) and are much less well defined in space, the optimal placement is rarely the one found in the original base placement,
but will be impacted more by the rest of the ligands placement.

The main scheme assumes a base fragment that has interaction points belonging to the same fragment. This is by far the fastest docking scheme.
Here we map the base fragment onto the earlier created conformers (Debug: mconformers.mol2),
followed by an optimisation of the conformers along the dihedrals of the non-pure single bonds that molden stored earlier.

However when a ligand triplet with two or more Hydrogen bond acceptor/donor
interaction points can not be formed from the same fragment, a different docking scheme is employed.

Now we use the whole of the conformations of the ligand to sample ligand points,
but to make the algoritm more efficient we only keep the unique ones
(unique by location of the interaction points origin) and store the conformations it belongs to with these interaction points.
After optimisation of the base fragments, the ligand poses are generated from the combined confomer information of the three interaction points.
If all three interaction points contain the same conformer, this conformer is mapped onto the base fragment,
followed by an optimisation of the conformers along the dihedrals of the non-pure single bonds that molden stored earlier.

The force field scoring is stored into the multi-mol result dock.mol2 file.
The source distribution of molden5.8 has a 'dock' subdirectory of the test directory, with three different docking:

• ligand dtst.xyz (part of MTX) into the 4dfr protein. (fastest: all three interaction points on the same fragment)
• ligand est.xyz into the 1a52 protein. Slower because the default interaction point separation of 10 angstrom has to be increased.
• ligand rsb.xyz (part of RAL(oxifen)) into the 1err protein. (slowest: NOT all three interaction points on the same fragment)

CAVEATS/Missing functionality:

• the orientation of the for example OH groups (Serine,Threonine,tyrosine) have to be predefined in the PDB file.
• Structural and Non-structural waters are ignored in the docking.
• Docking into an active site with metals is not yest supported.
• No ring conformations are being searched