popularly being used. The key differences among these tools are mostly in the
algorithms that are implemented for conformational search and alignment. This
chapter is about the general steps followed by most of the programmes to recognize
a pharmacophore pattern from a group of molecules that interact with a common
receptor and the diverse applications of the pharmacophore concept.
5.1.1 Picking the Right Set of Compounds and Their Initial Structures
As the resulting pharmacophore models are highly inclined by the type, size and
structural diversity of the participating ligands, it is imperative to choose the set of
ligands that take part in the process of pharmacophore model generation. Some
programmes like RAPID [69], HipHop [61] and the Crandell Smith method [70]
assume all the compounds in the set as active, some other methods consider the
information on the inactive molecules to be important as they give an idea about the
structural features responsible for reducing the activities and the ones essential for
enhancing activity. For example, DISCO [62, 71] and CLEW [72] provide an
option to include or exclude inactive molecules in generating a model so that the
user can identify the distinguishing features, while HypoGen [61] provides an
option for including activity ranges of the set of ligands. As far as size of the dataset
is concerned, most of the programmes are capable of handling up to 100 ligands in
a set. If the dataset contains large number of molecules, then it can be sorted and
categorized based on the activity value ranges. However, some programmes like
SCAMPI [73] can handle up to a few thousand molecules but compromising the
quality of the models. The high structural diversity of the dataset also is important
to identify features that are most essential for target binding and produce
high-quality models. Correct compound structures with correct atomic valencies,
bond orders and properly defined aromaticity and the appropriate stereochemical
flags are crucial for model generation.
5.1.2 Conformational Search
Ligands being flexible may have multiple possible conformations, and each conformation may bind to the binding site of the target in a particular fashion. Thus, it
is crucial to consider the flexibilities of each molecule during pharmacophore
development. Conformational search is considered as a separate stage in most of the
pharmacophore modelling programmes like HipHop, DISCO and RAPID, where a
large number of conformations are generated for each ligand. Systematic search,
Monte Carlo sampling and molecular dynamics are the methods of choice for most
of the software for conformation generation. As, the number of all possible conformers for molecules (especially when they have complex structures with a large
number of rotatable bonds) is too large to handle and incorporate in the pharmacophore model building, energy minimization and clustering methods are used to
Pharmacophore Modelling and Screening: Concepts, Recent …
33
algorithms that are implemented for conformational search and alignment. This
chapter is about the general steps followed by most of the programmes to recognize
a pharmacophore pattern from a group of molecules that interact with a common
receptor and the diverse applications of the pharmacophore concept.
5.1.1 Picking the Right Set of Compounds and Their Initial Structures
As the resulting pharmacophore models are highly inclined by the type, size and
structural diversity of the participating ligands, it is imperative to choose the set of
ligands that take part in the process of pharmacophore model generation. Some
programmes like RAPID [69], HipHop [61] and the Crandell Smith method [70]
assume all the compounds in the set as active, some other methods consider the
information on the inactive molecules to be important as they give an idea about the
structural features responsible for reducing the activities and the ones essential for
enhancing activity. For example, DISCO [62, 71] and CLEW [72] provide an
option to include or exclude inactive molecules in generating a model so that the
user can identify the distinguishing features, while HypoGen [61] provides an
option for including activity ranges of the set of ligands. As far as size of the dataset
is concerned, most of the programmes are capable of handling up to 100 ligands in
a set. If the dataset contains large number of molecules, then it can be sorted and
categorized based on the activity value ranges. However, some programmes like
SCAMPI [73] can handle up to a few thousand molecules but compromising the
quality of the models. The high structural diversity of the dataset also is important
to identify features that are most essential for target binding and produce
high-quality models. Correct compound structures with correct atomic valencies,
bond orders and properly defined aromaticity and the appropriate stereochemical
flags are crucial for model generation.
5.1.2 Conformational Search
Ligands being flexible may have multiple possible conformations, and each conformation may bind to the binding site of the target in a particular fashion. Thus, it
is crucial to consider the flexibilities of each molecule during pharmacophore
development. Conformational search is considered as a separate stage in most of the
pharmacophore modelling programmes like HipHop, DISCO and RAPID, where a
large number of conformations are generated for each ligand. Systematic search,
Monte Carlo sampling and molecular dynamics are the methods of choice for most
of the software for conformation generation. As, the number of all possible conformers for molecules (especially when they have complex structures with a large
number of rotatable bonds) is too large to handle and incorporate in the pharmacophore model building, energy minimization and clustering methods are used to
Pharmacophore Modelling and Screening: Concepts, Recent …
33