to aid protein structure refinement [159], proving useful in understanding interactions, in molecular geometry, in ligand fitting and in understanding overall protein
structure [160]. Combining more precise geometric and derived data from small
molecule structures in the CSD with PDB data can aid in the identification of
potential new drug molecules. The following three areas illustrate how the overlap
between small molecule and macromolecular crystallography can work together
through: analysis of interactions in binding sites, searching of ligand-protein complexes and generating likely conformational binding poses for new potential drug
molecules.
SuperStar [161] provides a knowledge-based approach to assist in identifying
interaction sites in proteins. It uses information from IsoStar [146], which contains
information about non-bonded interactions from both the CSD and PDB, to generate
interaction maps within protein binding sites or around small molecules, i.e. it
generates ‘hot-spots’ where a chosen interaction is particularly favourable.
Relibase [162] was a program building upon work published by Manfred
Hendlich in 1998, which provides a means of storing and analysing protein-ligand
complexes from the PDB. Query types included 2D substructure, 2D similarity, 3D
substructure and sequence similarity searching. Relibase+ [163] additionally provided extensions to the software for handling crystallographic packing effects. Later
publications provide further details of the design and curation of the Relibase
database [164] and its application in drug design [165].
Another way that CSD data has been utilised is in the program GOLD [166]
(Genetic Optimisation for Ligand Docking) which uses a genetic algorithm to
predict the binding modes of flexible ligands into protein binding sites, a problem
which is key in rational drug design. The docking of ligands is facilitated if they are
first optimised using the CSD Conformer Generator [144]. A library of ring conformations extracted from the CSD can also be utilised by GOLD. CSD data has been
used to parameterise other force fields, like COMPASS [167] and CHARMM [168],
as well as in external conformer generators such as CONFECT [169] and BCL::
CONF [170].
The value of being able to integrate different data sources has become clear in
recent years, and to this end, simultaneous searching of the CSD and PDB has been
integrated into CSD-CrossMiner [171]. This is a tool for pharmacophore-based
searches which grew out of a collaboration between CCDC and the ComputerAided Drug Design Section of F. Hoffmann-La Roche Ltd. at Roche Innovation
Center Basel. It provides interactive searching for, amongst others, protein-ligand
interaction patterns, ligand scaffolds or protein environments and can be used
concurrently to search protein-ligand binding sites from the PDB and small organic
molecules from the CSD using the same pharmacophore query. Searching just the
CSD for 3D features from the binding site can reveal potential molecules that may
bind to the protein. Reversing the search to look for proteins from the PDB may
highlight other proteins with a similar binding pocket with which the ligand could
also interact.
118
S. J. Coles et al.
structure [160]. Combining more precise geometric and derived data from small
molecule structures in the CSD with PDB data can aid in the identification of
potential new drug molecules. The following three areas illustrate how the overlap
between small molecule and macromolecular crystallography can work together
through: analysis of interactions in binding sites, searching of ligand-protein complexes and generating likely conformational binding poses for new potential drug
molecules.
SuperStar [161] provides a knowledge-based approach to assist in identifying
interaction sites in proteins. It uses information from IsoStar [146], which contains
information about non-bonded interactions from both the CSD and PDB, to generate
interaction maps within protein binding sites or around small molecules, i.e. it
generates ‘hot-spots’ where a chosen interaction is particularly favourable.
Relibase [162] was a program building upon work published by Manfred
Hendlich in 1998, which provides a means of storing and analysing protein-ligand
complexes from the PDB. Query types included 2D substructure, 2D similarity, 3D
substructure and sequence similarity searching. Relibase+ [163] additionally provided extensions to the software for handling crystallographic packing effects. Later
publications provide further details of the design and curation of the Relibase
database [164] and its application in drug design [165].
Another way that CSD data has been utilised is in the program GOLD [166]
(Genetic Optimisation for Ligand Docking) which uses a genetic algorithm to
predict the binding modes of flexible ligands into protein binding sites, a problem
which is key in rational drug design. The docking of ligands is facilitated if they are
first optimised using the CSD Conformer Generator [144]. A library of ring conformations extracted from the CSD can also be utilised by GOLD. CSD data has been
used to parameterise other force fields, like COMPASS [167] and CHARMM [168],
as well as in external conformer generators such as CONFECT [169] and BCL::
CONF [170].
The value of being able to integrate different data sources has become clear in
recent years, and to this end, simultaneous searching of the CSD and PDB has been
integrated into CSD-CrossMiner [171]. This is a tool for pharmacophore-based
searches which grew out of a collaboration between CCDC and the ComputerAided Drug Design Section of F. Hoffmann-La Roche Ltd. at Roche Innovation
Center Basel. It provides interactive searching for, amongst others, protein-ligand
interaction patterns, ligand scaffolds or protein environments and can be used
concurrently to search protein-ligand binding sites from the PDB and small organic
molecules from the CSD using the same pharmacophore query. Searching just the
CSD for 3D features from the binding site can reveal potential molecules that may
bind to the protein. Reversing the search to look for proteins from the PDB may
highlight other proteins with a similar binding pocket with which the ligand could
also interact.
118
S. J. Coles et al.