thermodynamics of water molecules. We provide the option for
the user to use the Reduce [19] program to adjust the proton-
ation state for HIS based on the local environment, and ASN
and GLN side chains may be flipped to optimize the hydrogen-
bond network. The user may also use other tools, such as
protein preparation wizard in maestro [24], for the proton-
ation states prediction. The hydrogens in ASP, GLU, HIS, and
LYS will be evaluated to give appropriate residue names com-
patible with amber force field.
- Selection of force field and water model—The current version
allows choosing from three different amber force fields.
Amber14SB [20] is the recommended choice in the latest
amber manual. Two other force fields, Amber99SB [25] and
Amber99SBildn [26], can also be chosen depending on the
user’s interest. Five water models with associated atomic ions
have been tested and implemented in the current version of
WATsite. TIP3P and SPC/E are three-site water models, and
the SPC/E model adds an average polarization correction to
the potential energy function [27]. TIP4P and TIP4P-Ew are
four-site water models which have four interaction points by
adding one dummy atom near the oxygen along the bisector of
the HOH angle. The TIP4P-Ew model was reparameterized
for use with Ewald summation methods [28]. OPC is a new
four-site and three-charge rigid water model which has quite
different point charges and charge–charge distances [29]. - Timestep and number of steps of simulation—Bond stretching
is a fast motion which determines the size of the MD timestep.
With bond length constraints applied on the bonds involving
hydrogen atoms, a timestep of 2 fs can be used. A time step of
equal or less than 1 fs is required if constraints are not applied.
With a 2 fs timestep, at least 2,000,000 steps for the produc-
tion simulation are required based on previous analysis
[23]. Hydration sites locations and free energies converged
over the 4 ns trajectory since all protein heavy atoms and
potentially present ligand heavy atoms were harmonically
restrained (spring constants are user-specified). - Clustering method—To identify locations of hydration sites in
high water density regions, a clustering method is used in
WATsite analysis. Currently, two clustering algorithms have
been implemented: quality threshold (QT) [30] and density-
based spatial clustering of applications with noise (DBSCAN)
[31]. QTclustering method is superior in identifying hydration
sites with a large occupancy value during the simulation,
whereas DBSCAN clustering method is advantageous in iden-
tifying more hydration sites with lower occupancy throughout
the simulation.
400 Ying Yang et al.