Identifying the ligand-binding site in the three-dimensional (3D) structure of a target protein is a starting point for the rational design of therapeutic agents. Many procedures have been developed for the detection of binding sites in protein 3D structures. These are divided into two types of algorithms: those based on geometrical cavity detection in a protein, and those based on the detection of interaction points on a protein. The first type includes a surface method (MS), 2,3) grid-surface methods (Cavity Search 4) and VOIDOO, 5) ) alpha-shape methods derived from the Voronoi diagram (CAST 6) and VOLBL 7) ), a layer method using a probe (PASS), 8) and a mapping method using the hydrophobic groups on a protein surface. 9) These procedures are often unable to detect binding sites close to the protein surface or to identify more open binding sites. The second type includes the detection of interaction points by a probe using grid searching methods , DOCK [14][15][16] , QSiteFinder 17) and SiteMap 18) ) and a random searching method (MCSS). [19][20][21][22] These procedures used several probes (methane, water, etc.) to detect pockets on a protein, and then evaluated likelihoods of the binding site with detected pockets by physical or empirical functions (van der Waals, electrostatic, hydrogen-bonding, etc.). However, because the functions of these interactions are short range, the representation of a binding site on a protein is discrete and does not correspond to the whole binding site.Hydrophobicity is a key parameter for understanding drug activity 23,24) ; however, it has rarely been used in the identification of binding sites, because it is difficult to define. Hydrophobicity is usually measured and reported as a log P value, where P is the partition coefficient of the molecules in octanol/water. However, the dependency of the hydrophobic interaction on distance has remained unknown, and it has therefore been approximated by various functions of distance. Hydrophobic-potential functions that use lipophilic constants based on log P are known as molecular lipophilicity potentials (MLPs). [25][26][27][28][29][30][31][32][33] Audry's first MLP was based on a hyperbolic distance function, Eq. 1.
25,26)Here Kellogg et al. proposed HINT (Hydrophobic INTeractions) function which is atom-based, Eq. 3.
29-32)MLPÏs i a i s j a j exp(ÏȘd ij )Here, hydrophobic atom constants, a i and a j , are multiplied by the solvent-accessible surface area of each atom, s i and s j . Gaillard et al. introduced the decay length of 2 Ă
into an exponential function, Eq. 4.
33)MLPÏf i f j exp(ÏȘd ij /2)All of MLPs were fitted to the log P values of compounds on the assumption that they were situated in contact positions or associated positions. Therefore, the short decay length was calculated as 1 or 2 Ă
(Eqs. 2-4). Israelachvili and Pashley measured the hydrophobic interaction between two monolayer-coated mica surfaces in aqueous solutions, and approximated the results using the exponential expression shown in Eqs. 5a-c.
34)Here,...