placement predictions or returned “in place” after the water
placement is accomplished. Then, the water molecules over-
lapping with the ions should be discarded.
- If an initial structure includes a ligand in the binding site of
interest, the unbound receptor structure should be minimized
after discarding the ligand to avoid the bias of using a prebound
crystal structure, which could otherwise be a source of bias/
error both for water placement and consequent docking
performance.
- Instead any other protocol for energy minimization as, for
example, an open source energy minimization implementation
in Open Babel [20] can be used.
- Depending on electrostatic properties of a particular receptor,
absolute values of the displayed isosurfaces corresponding to
similar-sized confined volumes could be very different making
the decision about the number of water molecules to place
arbitrary and ambiguous. Therefore, such a decision should
be qualitatively based on previous experimental data on the
average amount of water molecules in the protein–ligand inter-
faces of similar sizes/geometry/composition and on sizes of
the volumes corresponding to the chosen isovalues that should
be big enough to cover at least one water molecule.
- As inNote 5, this step should be carried out manually and
cannot be strictly defined qualitatively a priori. Calibration
studies on the experimental structures of the complexes similar
to the studied one should be carried out in order to gain
qualitative insights about how hydrophobic/hydrophilic are
the interactions in this particular class of protein–ligand
interfaces.
- Due to the qualitative nature of such predictions, it is recom-
mended to carry out molecular docking both with and without
water molecules included into the structure of the receptor and
to compare the results rigorously to find out the putative role
of the particular water molecules in the interface. Consequent
application of MD simulations to the obtained complexes
could also help to understand how properly the positions of
water molecules are predicted.
- In case no information about the binding region is available,
protein surface can be split into several parts, for each of which
an independent DMD study could be carried out. Then, the
docking poses obtained for these parts should be compared in
terms of scoring to derive the best candidates for binding poses.
Since such a division of the protein surface in several segments
is not unique, several combinations of segmentation (for exam-
ple, obtained by shifting the segments representing putative
Solvent Inclusion in Docking Glycosaminoglycans 451
