Comparative Analysis of Techniques Oriented on the Recognition of Ligand Binding Area in Proteins - Identification of Ligand Binding Site and Protein-Protein Interaction Area - page 58

4.4 Tools Under Consideration

4.4.1 Geometry-Based Techniques

4.4.1.1 CASTp - Computed Atlas of Surface Topography

of Proteins ( http://sts.bioengr.uic.edu/castp )

CASTp identifies ligand binding sites by studying the geometric properties of
protein pockets under the assumption that ligands are naturally attracted to depressions
in the protein body. Thus, its core algorithm searches the 3D representation of the
protein for pockets capable of housing a solvent molecule with a diameter of 1.4 Ǻ.
The authors refer to such pockets as “mouths”. In contrast, a cavity is a depression
which remains inaccessible to the solvent molecule and therefore does not have the
properties of a “mouth”. Identification of binding sites bases on a computational
geometry model capable of locating “pockets” and “cavities”. Cavity determination
parameters are not dependent on the rotation of the molecule. Moreover, CASTp
does not employ grid coordinate analysis (Liang et al. 1998a ) . Cavities are identi fi ed
by way of weighted Delaunay triangulation and applying the alpha complex for
shape measurements (Edelsbrunner and Mucke 1994 ; Edelsbrunner 1995 ; Facello
1995 ; Edelsbrunner and Shah 1996 ; Edelsbrunner et al. 1998 , 1995 ). These
methods return the surface of the accessible pockets as well as of internal (inacces-
sible) cavities. For each cavity the program calculates its area, volume and sol-
vent-accessible surface (with respect to molecular surface).
In our analysis from among the files listing atoms representing all detected pockets
and cavities, the extracted amino acids corresponding to a single, specific protein
chain were selected and then compared with the reference database (PDBSum - used
as the golden standard).
A detailed discussion of CASTp algorithms can be found in Liang et al. ( 1998a, b );
Binkowski et al. ( 2003 ) and Dundas et al. ( 2006 ).

4.4.1.2 Pocket-Finder ( http://www.modelling.leeds.ac.uk/pocket fi nder/ )

Pocket-Finder is an extension of an existing software package called Ligsite, developed
by Hendlich et al. ( 1997 ). It uses a grid system with a resolution of 0.9 Ǻ, centered
upon the target protein. A scanning probe with a radius of 1.6 Ǻ along each axis is
used. This probe also enables testing cubic diagonals. Identifying a pocket requires
locating an area where a grid point which belongs to the protein molecule is adja-
cent to a grid point which represents empty space, which is itself adjacent to another
protein-bound point. Identification of the status of each grid point is performed in
seven directions, resulting in seven separate results for each point. If five of these
results are positive, the empty space is treated as a cavity.
PDB-derived proteins are scanned for ligands. If the contact between molecule
and protein is possible, this molecule is treated as possible ligand.