SILCS-Pharm: Receptor-Based Pharmacophore Models from FragMaps
==============================================================

Background
----------

Three-dimensional (3-D) pharmacophore models (also called "hypotheses") offer an
intuitive and powerful approach for virtual screening (VS). 3-D pharmacophore VS
works by searching for matches between the 3-D pattern of functional groups
constituting a pharmacophore model and the 3-D arrangement of functional groups
in ligand conformers in a virtual database. Ligands having conformers that
closely match the pharmacophore model are considered hits. Traditionally, 3-D
pharmacophore models were developed using experimental knowledge of the binding
poses for ligands to a receptor. This is in contrast to energy function-based VS
("docking"). An advantage of docking approaches was that they only require
experimental knowledge of the receptor structure, and not of bound ligands.
However, it is possible to develop pharmacophore models using only the receptor
structure. One means to generate such receptor-based pharmacophore models is
with data from SILCS simulations.


SILCS-Pharm converts Grid Free Energy (GFE) FragMaps into 3-D pharmacorphore
models. GFE FragMaps from SILCS simulations are used as inputs for the four-step
SILCS-Pharm process:

1. voxel selection; 
2. voxel clustering and FragMap feature generation; 
3. FragMap feature to pharmacophore feature conversion; 
4. generation of a pharmacophore model for virtual screening (VS).

Here, "feature" refers to the identity and location of a chemical functional
group and "pharmacophore model" is a collection of pharmacophore features.

FragMap generation from SILCS MD entails partitioning 3-D space into uniform
cubic voxels and enumerating fragment binding probabilities for each voxel. The
voxels are retained during Boltzmann-inversion of probabilities to create GFE
FragMaps. As the GFE voxel value represents the binding strength of a functional
group at that specific location on the protein surface, the first step
identifies the most favorable interactions based on a particular GFE cutoff
value. In the second step, clustering is performed to group adjacent voxels,
with each unique cluster becoming a FragMap feature. In the third step, the
FragMap features are classified and converted into pharmacophore features.
Finally, the pharmacophore features are prioritized using Feature GFE (FGFE)
scores to create a SILCS 3-D pharmacophore model :cite:`Yu:2015d`. In the
current scheme, the program searches through five FragMap types: Generic Apolar,
Generic Donor, Generic Acceptor, Methylammonium N, and Acetate O. Additionally,
instead of using a rigidly-defined protein surface to determine regions that
ligands cannot sample because of protein volume occlusion, the SILCS Exclusion
Map is used. The Exclusion Map has been validated to be a better alternative to
more traditional representations of the occluded volume of the protein, and it
takes protein flexibility into account in an explicit way.

Usage and Options
-----------------

A single Command Line Interface command performs the four-step SILCS-Pharm 
process:

::

   ${SILCSBIODIR}/silcs-pharm/1_calc_silcs_pharm prot=<prot pdb> center="x,y,z"

The input arguments are the PDB file used for the SILCS run and the absolute
position of the center of a 10 Å sphere to be used to define the boundaries of
the pharmacophore model. Two output files result from this command.
``<prot>.keyf_<#features>.ph4`` can be directly used for 3-D pharmacophore VS by
compatible programs (see below for generating Pharmer-compatible ph4 files).
``<prot>_silcspharm_features.pdb`` provides output in PDB format for easy
visualization using standard molecular graphics packages. Running the command
with no arguments

::

   ${SILCSBIODIR}/silcs-pharm/1_calc_silcs_pharm

will list additional options. Along with the required arguments of ``prot=<prot
pdb>`` and ``center="x,y,z"``, the following additional options can be set:

1) FragMap directory path: 

   ::
  
      mapsdir=<location and name of directory containing FragMaps; default=2b_gen_maps>

   By default, the program looks for FragMaps in the ``2b_gen_maps`` directory.

2) Output directory path: 

   ::
  
      outputdir=<location and name of output directory; default=5_pharm>

   By default, the program creates the directory ``5_pharm`` and places all
   output files there.

3) Radius: 

   ::

      radius=<default: 10>

   By default, a radius of 10 Å centered at center="x,y,z" is searched to
   generate pharmacophore features from the input FragMaps.

4) Generic Apolar FragMap cutoff:

   ::

      apolar_cutoff=<default: -1.2>

   By default, Generic Apolar FragMap voxels having a GFE value <= -1.2 kcal/mol
   are selected.

5) Generic Donor FragMap cutoff :

   ::

      hbdon_cutoff=<default: -1.0>

    By default, Generic Donor FragMap voxels having a GFE value <= -1.0 kcal/mol
    are selected.

6) Generic Acceptor FragMap cutoff :

   ::

      hbacc_cutoff=<default: -1.0>

   By default, Generic Acceptor FragMap voxels having a GFE value <= -1.0
   kcal/mol are selected.

7) Methylammonium N FragMap cutoff :

   ::

      mamn_cutoff=<default: -1.8>

   By default, Methylammonium N FragMap FragMap voxels having a GFE value <=
   -1.8 kcal/mol are selected.

8) Acetate O FragMap cutoff :

   ::

      aceo_cutoff=<default: -1.8>

   By default, Acetate O FragMap FragMap voxels having a GFE value <= -1.8
   kcal/mol are selected.

   In addition to visualization, the output PDB file
   ``<prot>_silcspharm_features.pdb`` can be used for easy editing of the
   pharmacophore model. Using a text editor, modify/reduce the features in this
   file as desired and save the revised file as
   ``<prot>_silcspharm_features_revise.pdb``. Using this revised PDB file and
   the original ph4 file ``<prot>.keyf_<#features>.ph4`` as input, create a new
   ph4 file with:

   ::

      ${SILCSBIODIR}/programs/revise_ph4 <prot>.keyf_<#features>.ph4 <prot>_silcspharm_features_revise.pdb

   The output of this command will be the revised ph4 file
   ``<prot>.keyf_<#features>_revise.ph4``. To create Pharmer-compatible ph4
   output, add the ``pharmer`` option:

   ::
   
      ${SILCSBIODIR}/programs/revise_ph4 <prot>.keyf_<#features>.ph4 <prot>_silcspharm_features_revise.pdb pharmer

Example
-------

The following example demonstrates use of SILCS-Pharm to generate a
pharmacophore model for p38 MAP kinase. Input files, including FragMaps, are
provided in ``${SILCSBIODIR}/examples/silcs/``.

The x,y,z coordinates of the center of the complexed ligand are used to create a
SILCS-Pharm 3-D pharmacophore model encompassing the ligand binding site (shown
below). These coordinates define the center of the sphere within which FragMap
voxels are searched and clustered to generate features.

.. figure:: ../images/silcs-pharm/p38_pocket_maps.png
   :width: 90%

Using the ligand center coordinates ``center="35.24, 27.48, 37.73"``, generate
the 3-D pharmacophore model with:

::

   ${SILCSBIODIR}/silcs-pharm/1_calc_silcs_pharm prot=3fly.pdb center="35.24,27.48,37.73" mapsdir=${SILCSBIODIR}/examples/silcs/silcs_fragmaps_3fly/maps

The command will complete after several seconds, and the output will note, "A
total of 13 features have been detected." All output files will be in a new
subdirectory ``5_pharm``. Standard molecular graphics software can be used to
visualize the output file ``3fly_silcspharm_features.pdb``, which has one
``ATOM`` entry for each of the 13 pharmacophore features:

.. figure:: ../images/silcs-pharm/silcs_pharm_features.png
   :width: 90%

   Conversion of FragMaps (left) to SILCS pharmacophore features (right).

To modify/reduce the features, edit the output file
``3fly_silcspharm_features.pdb`` and save it as
``3fly_silcspharm_features_revise.pdb``, then run
``${SILCSBIODIR}/programs/revise_ph4``:

::

   ${SILCSBIODIR}/programs/revise_ph4 3fly.keyf_13.ph4 3fly_silcspharm_features_revise.pdb

The output ``3fly.keyf_13_revise.ph4`` will reflect your revisions in
``3fly_silcspharm_features_revise.pdb``.

.. figure:: ../images/silcs-pharm/p38_pocket_features_modify.png
   :width: 40%

   Revised SILCS pharmacophore features.