Static approaches

There are various rapid methods for prediction of hydration sites on molecular surfaces or in interfaces. A common feature of these methods is that they are focused on the protein molecule or on protein-water interactions and completely neglect water-water interactions and co-operations. Notably, these interactions largely determine residence of water molecules in the hydration network. Most rapid methods use a static picture disregarding dynamic exchange (mobility) between surface and bulk water molecules and focusing on protein(ligand)-water interactions. Prominent examples of static algorithms are described below.

Knowledge-based. Using structural reference data sets distilled from crystal structures of the PDB early methods were published for detection of hydration sites (Pitt and Goodfellow 1991).

Structural. It was shown that e.g. directionality of hydrogen bonds can be applied for systematic solvation of proteins (Vedani and Huhta 1991).

Scoring. A method based on docking of water molecules to the protein binding sites and subsequent use of a scoring scheme for the selection process was recently introduced (Ross et al. 2012).

Thermodynamics. A force field-based approach (Schymkowitz et al. 2005) used free energy calculations in combination with the knowledge-based method of Pitt and Goodfellow (1991). A statistical mechanics-based approach with Monte Carlosampling of possible hydration site configurations (Michel et al. 2009) was also developed. The method starts with definition of the binding site and fills up a grid covering the site with water molecules. Dynamic exchange of water molecules between the binding site and the bulk is not performed explicitly: an idealized particle concept is used to calculate exchange thermodynamics between bulk and the site.


Dynamic approaches

Molecular dynamics (MD) has long been applied (Rossky and Karplus 1979, van Gunsteren et al. 1983, Pettitt and Karplus 1987) for investigation of hydration of peptides and proteins. All-atom MD with explicit water models is an invaluable source of mobility information of any hydrated biological systems. During the simulation time, movements (trajectory) and all interactions of water molecules can be followed at atomic level including not only protein-water, but also water-water contacts and exchanges of primary importance. Two main branches of approaches applying MD for prediction of hydrate structure are discussed below.

Density-based calculations. Several studies (Virtanen et al. 2010, Makarov et al. 1998, Lounnas et al. 1994) have dealt with MD-based calculation of average solvent density and construction of proximal radial distribution function (pRDF) of hydration shells for different atom types occurring in proteins. The aim of these studies is to use the constructed, generalized pRDFs for the reconstruction of hydration shell density of any protein without MD simulation. In other words, this approach applies MD for calculation of solvent density and construction of pRDFs. Positions of individual water molecules can be obtained from fits to densities. Limitations of the radial distribution function-based approaches were discussed in details (Henchman and McCammon 2002).

Occupancy-based calculations. With advancement of computational infrastructure and theory speed of MD calculation have increased in the past decades (Dror et al. 2012). It has become a real alternative to perform atomic level MD with explicit water molecules for analysis (Schoenborn et al. 1995) and direct prediction of hydration structure of a protein or its complex. Whereas there are numerous analysis studies, there are much fewer studies on testing the usefulness of direct MD approaches for obtaining hydration sites (Huang et al. 2008, Henchman and McCammon 2002, Madhusudhan and Vishveshwara 2001). Direct MD approaches use individual positions of hydrating water molecules (instead of average densities) and apply various occupancy-based evaluation schemes to obtain hydration sites. For example, Henchman and McCammon (2002) define time averaged positions for this purpose. MobyWat also works with occupancy values and uses water mobility for prediction or analysis of the hydration structure.