Ligands with similar pharmacohoric patterns can exhibit similar biological activities
according to one specific protein. Thus, this chapter gives a few guide lines how to model a pharmacophore and
find ligands of a molecule library, that match to that specific pharmacophore. You can use the
virtual screening tool in
LigandScout
or through the
command line
.
Create a pharmacophore model (
structure-based
or
ligand-based
),
according to a target protein, that is capable to find a high amount of
active ligands and as few as possible inactives.
Here we refer a structure-based pharmacophore of 1ke7 complex as an example. One of the essential
steps is an exhaustive literature search about your protein of interest. Where does the
ligand bind? Which amino acids are required such that the ligand can interact with protein's
binding pocket. Thus, this leads to the knowledge about features that should be contained
in the pharmacophore. Which ligands bind to the protein and show biological activity
and which do not? Load your protein-ligand complex in the
Structure-Based Modeling Perspective
, zoom into the active site, check
the double bonds of the ligand and create a pharmacophore. Check if the essential interactions
are present and
edit
the pharmacophoric features if required.
It might be that several pdb complexes are available and therefore many
pharmacophore models can be generated. In the
Alignment Perspective
you can
edit the pharmacophores, align them, generate a shared or merged feature pharmacophore.
Just add your pharmacophores and ligands one by one to the Alignment Perspective and investigate them until
you have created an appropriate pharmacophore.
See also
Chapter 6,
Structure-Based Pharmacophore Design
,
Chapter 7,
Pharmacophore Alignment
and
Chapter 8,
Ligand-Based Pharmacophore Design
for further information on pharmacophore creation.
LigandScout allows you to screen single or multi-conformational molecule databases.
To efficiently speed up the screening process, LigandScout uses the proprietary LDB file format
for storing molecule libraries. Hence, it is necessary to convert your
molecule library into the LDB file format before you load your library for the screening.
LigandScout offers several ways to generate a multi-conformational database for screening.
If you have existing molecules available in LigandScout, you can export them via menu
File
>
Save as File
and choose the LDB file type.
Alternatively, you can generate a new screening library from an input file via the
Create Screening Database
icon, menu
Library
>
Generate Library from
, the command line tool
idbgen
or the stand-alone idbgen-GUI. .
For instance, store an existing multi-conformational library
screeningDB.sdf
as
screeningDB.ldb
use the following command:
idbgen --input screeningDB.sdf --output screeningDB.ldb --confgen-type import
Another idbgen command example shows how to generate from a SMI file
input.smi
a multi-conformational database
output.ldb
:
idbgen --input input.smi --output output.ldb --confgen-type omega-fast
Add your pharmacophore model to the
Screening Perspective
, e.g., by the
Copyboard Widget
.
Switch to the Screening Perspective and load a molecule library (.ldb) by clicking the
Load Database
button.
Then, choose a pharmacophore. LigandScout offers you to select more than one
database to be screened against the selected pharmacophore. Select and mark the database
containing active ligands in green. If a decoy database is available
that includes inactive ligands, mark it in red.
To customize the advanced screening settings, toggle on the
Show Advanced Options Settings
in the
Screening Panel
.
Then, start the screening process by pressing the
Perform Screening
button.
Optionally, you can screen a pharmacophore against an external
database in the Structure-Based Modeling Perspective. Make sure, that one pharmacophore is visible in
the 3D View and select in the menu
Pharmacophore
>
Screen Pharmacophore Against External Library
.
A
screening settings
dialog appears. Select your molecule database and modify the settings
as required.
To execute the virtual screening process through the command line,
please see
the section called “Virtual Screening”
in the Appendix.
When the screening process is finished, a hit list of molecules, that
matches the pharmacophore is shown in the
Library View
. Click the results one by one
and look how the ligands are aligned to the pharmacophore in the 3D View.
If you want to
export
your screening hits, select in the menu
File
>
Save as File
.
To see the ROC plot, press the
Plot ROC Curve
icon.
The ROC curve shows how good the pharmacophore model discriminates
between active and inactive ligands. The AUC value (area under the ROC curve)
quantifies the ROC curve between 0 and 1. A value near to 1 is favorable.
If you want to export the ROC plot as image, select in the menu
File
>
Save as File
. Choose an appropriate image file format (e.g. PNG) and
enter a file name in the
Save as
dialog. Then, select
the ROC check box plot of the 3D export properties.
If the screening results produce less active hits, there can be several reasons.
Question yourself, if the library ligands violate excluded volumes or
is the pharmacophore too restrictive? Are the conformers of
the ligands meaningful? Are essential features found in the ligands?
To get more hits,
edit
your pharmacophore model. For instance,
remove an excluded volume or other features or make the tolerance sphere larger and screen
again. Another way is to raise the
number of maximum omitted features
, e.g.,
1 or 2, depending on the size of the pharmacophore or designate some features as optional.
To do so, not all of the resulting ligand hits may contain all of the features of the
pharmacophore model, but again these can lead to interesting results. If you have the problem that too
many hits are found, then the model is too general and you should restrict your
pharmacophore model. You can create
excluded volumes
or an
excluded volumes coat
.
An excluded volumes coat consists of many small excluded volumes
which reflect the spatial geometry of the binding pocket.
Note that the combination of generalizing and restricting the pharmacophore
model is essential for influencing the hit results.
After all, finding a representative pharmacophore model
of a ligand-protein binding mode needs a lot validation. One approach
is to include active and inactive ligands in the molecule library. If
most of the active ligands are contained in the hit results, then the
pharmacophore covers the actives molecules well. On the other hand,
if inactive molecules are matched, then the pharmacophore might
possess the main feature scaffold matching the active ones; however
pharmacophore model could not differentiate between the active and
inactive ligands. In this case you will need to
build an improved model.