Ligands which bind to one specific target protein,
possess a similar binding mode. The goal of the ligand-based pharmacophore
modeling is to seek common features in a set of ligands. LigandScout provides
an automated generation of ligand-based pharmacophore models in the
Ligand-Based Modeling Perspective
or
command line
.
A vast literature search of your target protein and its
binding ligands are required. Although, if there are no protein structures
available, the information about essential amino acids for the binding mode
gives hints for verification of the results afterwards.
Gather known active ligands (at least 2) that bind to your target protein. This group of ligands serves
as a Training-Set for the main pharmacophore generation process.
These molecules are overlapped according to their pharmacophoric features
and then several pharmacophore models are created that are found in all
Training-Set molecules. A similar feature scaffold is one indication for this
group.
For instance, you can load in the Structure-Based Modeling Perspective 1ke6, 1ke7 and 1ke8 one by
one and add the ligands to the
Ligand-Based Modeling Perspective
by the
Copyboard Widget
.
In case you have an externally prepared ligands or Ligand-Set,
just load it into the Ligand-Based Modeling Perspective by the menu
File
>
Open
or alternatively, use the
Ligand-Set
>
Add Molecules: To Training-Set/Test-Set
.
Optionally, you can prepare a Test-Set, that includes actives and inactives.
This set is used for the verifying step, where the resulting pharmacophores are
aligned to the Test-Set molecules. Ligands which you do not want to include in
the process, can be put to the Ignored Ligands set. The type of the ligands can
be changed through the menu
Ligand-Set
>
Flag Molecule As
or switching the type in the Ligand-Set Table.
If you are not sure how to divide your ligands into Test-Set
and Training-Set, then LigandScout's clustering method assists your
choice. The aim of this clustering is to choose those compounds that are
similar in terms of 3D pharmacophore characteristics and therefore bear a
higher chance for delivering a large overlap of chemical features. The 3D
clustering algorithms performs fast alignments and clusters based on a similarity
value between 0 and 1. Since this algorithm basically performs combinatorial
alignments of all conformations of all compounds, a low number of conformations
(1-3) is recommended. The cluster distance can be varied until the desired
cluster size has been reached. To use the clustering method, select the
Clustering
icon.
Adapt the settings for clustering and conformation generation (if no conformers are
available yet). Then, click the OK button to run the clustering process.
A new column
Cluster ID
appears in the
Ligand-Set Table
.
The values show the group which the ligands are assigned to.
In the Ligand-Set Table, go through the ligands of each group to decide which type
of set you want to use for each ligand.
To generate conformations for your ligands press the
Generate Conformations for Ligand-Set
icon. A dialog appears where you can customize the
settings
.
Alternatively, you can generate conformations
through the
command line
.
When the ligand-set is ready to use for the pharmacophore generation,
just press the
Run Ligand-Based Pharmacophore Creation
icon.
A
settings
dialog appears where
you can adjust the properties for the process.
To understand how the settings influence the pharmacophore creation,
the process is explained shortly. First, conformations of the Training-Set
molecules are generated (if not available). After ranking the
molecules according to their number of conformations (flexibility), pharmacophore features
(lipophilic points, hydrogen bond donors and acceptors, positive and negative ionizable groups)
are projected on these molecules and all their conformations. All conformations of the two top
ranked (i.e. the least flexible) molecules are then aligned using Inte:Ligand’s
molecular alignment algorithm. For a configurable number of best alignment solutions
(“
intermediate alignment solutions
”) common pharmacophoric features are interpolated
and intermediate pharmacophore models are created and stored for further processing
(“
intermediate pharmacophore models
”). These intermediate pharmacophore models are
now ranked using several adjustable scoring functions taking into account chemical feature
overlap, steric overlap, or both. The intermediate pharmacophore models are then
aligned to all conformations of the third molecule, etc., and a new set of intermediate combined feature
pharmacophores is created until all molecules have been processed. If at any stage no conformation can be
found that can be matched on any intermediate solution, the process is stopped. If at least three common
chemical features can be identified throughout the whole alignment and interpolation process, the feature
pharmacophore combination is considered to be successful. All these steps can be logged in a file verbosely.
The output can either be written in LigandScout’s internal file format (PML or PMZ)
or e.g. in Catalyst/Discovery Studio hypoedit format for use in Catalyst.
If the pharmacophore is exported to Catalyst, one must take
into account that Catalyst is unable to map two features on one single
ligand atom - in this case the Catalyst hypothesis can be reduced by deleting features or
create several hypotheses for sequential virtual screening.
For forming intermediate solutions, two modes for sequential pairwise alignment
are provided: shared and merged feature pharmacophore combination mode. While with shared feature combination mode an
intersection between all pharmacophoric features of the underlying molecules is built (logical AND), with
merged feature pharmacophore building mode, the features of the aligned molecules will be merged (logical OR).
To avoid huge pharmacophores from merged feature pharmacophores, a parameter “
number of omitted features
”
must be specified. This parameter defines how many features per molecule may be omitted by the
algorithm with respect to the final pharmacophore solution. Specifying a value of “
1
” for this parameter makes
sure that a maximum of one feature is not matched by each of the Training-Set molecules.
If you want to generate exclusion volumes to the ligand-based pharmacophore model, just toggle on the
Create Excluded Volume
check box in the Ligand-Based Pharmacophore Creation dialog.
To sterical limit the pharmacophore, excluded volumes are placed around the best alignment model.
For more information about the settings, please see
the section called “Ligand-Based Modeling Settings”
and
the section called “Scoring Function”
.
After the ligand-based pharmacophore generation is finished, the results
are listed in the
Results Table
. Select the first result in the table. In the Hierarchy View
make all ligands visible by selecting the
Toggle Visibility
icon of each ligand.
Thus, you can see how well the resulting pharmacophore fits to the ligands in the 3D View.
In the Ligand-Set Table the
Feature Pattern
column shows the
color encoding of the features, that matches the ligands. There, you can see which features
cover the ligands well.