MPI and Multi-threading mix Parallel AutoDock Vina.

1. Compile the program.

1.1 MPI and Boost libraries are required for VinaLC.

VinaLC depends on two external libraries: MPI (https://www.open-mpi.org or https://www.mpich.org) and boost (https://www.boost.org).

Please install the MPI first. Users can use open-mpi or mpich. Either one is OK. Then install the boost. You need to enable the boost-mpi build. In the project-config.jam add one line to the end of the file:

using mpi : <path_to_your_MPI_installation_directory>/bin/mpicxx ;

to install Boost library please follow the step in the Boost document.

./bootstrap.sh --prefix=path/to/installation/prefix
./b2 install

On the LC machines, the MPI and Boost are installed. Users can source the environment and then compile the code.

On quartz:

module load boost/1.62.0

1.2 Obtain the code

The code can be download from:

https://github.com/XiaohuaZhangLLNL/VinaLC

by git:

git clone [email protected]:XiaohuaZhangLLNL/VinaLC.git

1.3 Installation

Installing it by using cmake is straight forward:

After you install or source the environment variables for both MPI and boost, run the cmake in the home directory of VinaLC:

cd <Home_Directory_VinaLC>
mkdir build
cd build
ccmake ../ -DCMAKE_INSTALL_PREFIX=<path_where_you_want_to_install_VinaLC> -DBOOST_ROOT=<path_to_boost_installation>  -DMPI_CXX_COMPILER=<path_to_your_MPI_installation_directory>/bin/mpicxx

make
make install

2. Input files.

In the vinalc/examples, there is a small test case contains input files:

2.1 receptor

file contains list of receptor file name: recList.txt two receptor pdbqt files:

1KIJ_protH.pdbqt 1KIJ_protH1.pdbqt

2.2 ligand

file contains list of ligand file name: ligList.txt a "data" directory contains two ligand pdbqt files:

data/
ligands1.pdbqt  ligands2.pdbqt

2.3 geometry file to save the docking grid information.

each line is coresponding to each receptor in recList.txt. each line has six number where first 3 are center of active site and last 3 are grids dimension. the file is arranged as:

x_center  y_center  z_center  x_grid  y_grid  z_grid
x_center  y_center  z_center  x_grid  y_grid  z_grid
....

2.4 The pdbqt files for receptors and ligands are prepared from their pdb file by mgltools

(http://mgltools.scripps.edu/) Two python scripts in MGLTools-1.5.6rc2/MGLToolsPckgs/AutoDockTools/Utilities24/ are used

for receptor

prepare_receptor4.py -r rec.pdb -o rec.pdbqt -A checkhydrogens

for ligand

prepare_ligand4.py -l ligand.pdb

3. Running program

3.1 To run the program with slurm in debug mode:

srun -N4 -n4 -c12 -ppdebug ./vina --recList recList.txt --ligList ligList.txt --geoList geoList.txt

-N4: 4 nodes will use
-n4: 4 tasks will  each task run on one node
-c12: 12 threads will run on each node
-ppdebug: use debug mode

3.2 Vina program option:

[zhang30@quartz1538 bin]$ ./vinalc --help

Input:
  --recList arg               receptor list file
  --fleList arg               flex part receptor list file
  --ligList arg               ligand list file
  --geoList arg               receptor geometry file
  --exhaustiveness arg (=8)   exhaustiveness (default value 8) of the global
                              search (roughly proportional to time): 1+
  --granularity arg (=0.375)  the granularity of grids (default value 0.375)
  --num_modes arg (=9)        maximum number (default value 9) of binding modes
                              to generate
  --seed arg                  explicit random seed
  --randomize                 Use different random seeds for complex
  --energy_range arg (=2)     maximum energy difference (default value 2.0)
                              between the best binding mode and the worst one
                              displayed (kcal/mol)
  --useScoreCF                Use score cutoff to save ligand with top score
                              higher than certain critical value
  --scoreCF arg (=-8)         Score cutoff to save ligand with top score higher
                              than certain value (default -8.0)

Information (optional):
  --help                      display usage summary

Error: Total process less than 2

3.3 Run with different options

srun -N4 -n4 -c12 ./vina --recList recList.txt --ligList ligList.txt --geoList geoList.txt --exhaustiveness 12
srun -N4 -n4 -c12 ./vina --recList recList.txt --ligList ligList.txt --geoList geoList.txt --exhaustiveness 12 --granularity 0.333
...

options, --recList --ligList  --geoList must be specified  

4. The programs to prepare the input PDBQT file.

4.1 babel

If you start with SDF files for the ligands, you can use the babel to convert the SDF to PDBQT directly:

babel -isdf <ligand-file-name>.sdf -opdbqt <ligand-file-name>.pdbqt

4.2 MGLTools(http://mgltools.scripps.edu/)

The AutoDock developer team provides graphic user interface, AutoDockTools (ADT), to prepare the input files. The receptor input file MUST use ADT to convert the file format.

There is a Vina video tutorial to show how to use ADT to prepare receptor, ligand, and determine the grid size that use in the program.

http://vina.scripps.edu/tutorial.html

An important thing to remember when calculate the grid size:

x_grid=<number of poiont in x-dimension>*spacing
y_grid=<number of poiont in y-dimension>*spacing
z_grid=<number of poiont in z-dimension>*spacing

spacing in ADT is equal to granularity in Vina.