General Introduction of SPANA
SPANA adopts MPI/openMP hybrid parallelization scheme. In this scheme, a boundary is set for each snapshot structure you want to analyze. Generally, the dimension of this boundary is equal to that of the Periodic Boundary Condition (PBC) box. In the MPI parallelization, the boundary is decomposed into smaller domains based on parameters “domain_[x, y, z]” specified in [BOUNDARY] section. Each domain is further decomposed into cells with parameters “num_cell_[x, y ,z] in [BOUNDARY] section. Each domain has buffer region with thickness specified by a parameter “buffer” in [SPANA_OPTION] section. Here, we call the domain+buffer region ‘sub_boundary’. Each MPI process shares the same molecule information defined in a psf file or a prmtop file and in a PDB file
. However, each MPI process selectively assigns atoms located inside the sub_boundary into cells. This approach reduces memory usage required for cell information, and enables each MPI process to get the atom information (such as serial index and xyz coordinates) in target cells in the sub_boundary. Calculation for each domains is done by each MPI process, and calculation for cells are decomposed into OpenMP threads.
When [BOUNDARY] type = PBC and [SPANA_OPTION] wrap =yes, all atoms are wrapped into the unit PBC box using the box size information written in the trajectory file (dcd file). If the file does not have the box size information, the option ‘wrap = yes’ does not work. When wrap = no, the program does nothing for atomic coordination. Some SPANA application (e.g. “rdf_analysis”) has an option ‘recenter’. The vector from the center of the unit box to the center of mass of the specified group is calculated, and all atoms are translated using this vector. In this case, wrapping of the atomic coordinates is performed after the centering. When you use the option ‘recenter’, please make sure that the target solute molecule is not chopped with PBC boundary. If recenter = 0, or the ‘recenter’ option does not appear in a control file, the centering is not done. Some SPANA application (e.g., density_analysis) uses fitting options in [FITTING] section. In this case, it is recommended to recenter the group for fitting. The fitting will be performed after the recentering and wrap. Note that after the fitting, atoms outside of the boundary are not wrapped again, and are ignored by analysis.
The boundary of the whole system is determined by option ‘box_size’ in [SPANA_OPTION] section. In the case of [BOUNDARY] type= PBC and box_size = trajectory, the boundary is determined from the box information written in the trajectory file. Therefore, the boundary size is the same as that of PBC box. On the other hand, when box_size = manual, the boundary size is determined from the values box_size_[x, y, z] in [BOUNDARY] section. This option is useful to save the memory usage, because, when this option is used, the program performs a calculation only for the selected region around the target solute molecule. In this case, atoms outside of the manually determined boundary are not assigned into any cells and ignored by analysis. Please note that the target molecule should be centered when you use box_size = manual.
When [BOUNDARY] type = NOBC, the boundary used for the analysis should be determined with box_size = manual.
General example of SPANA input file
The following sample is the typical input for the SPANA. This input is used to calculate the radial distribution function of hydration water (see rdf_analysis for detail) around BPTI (group1) in water box with PBC. The whole system is decomposed into 8 domains by domain_[x, y, z] = 2. Furthermore, the whole system is decomposed into 136 x 160 x 132 cells by num_cell_[x, y, z]. The boundary size is determined by PBC box size written in the trajectory file. In every frame for the analysis, coordinates of all atoms are translated by the vector used for the centering of BPTI (recenter = 1). After the translation, all atoms are wrapped (wrap=yes) inside the PBC box. The thickness of the buffer region (i.e., thickness of the sub_boundary around domain) is set to 10 by the option cutoff.
[INPUT] psffile = ionize.psf reffile = ionize.pdb pdbfile = ionize.pdb [OUTPUT] txtfile = bpti_proximal.out [TRAJECTORY] trjfile1 = run.dcd md_step1 = 500000 mdout_period1 = 500 ana_period1 = 5000 repeat1 = 1 trj_format = DCD # (PDB/DCD) trj_type = COOR+BOX # (COOR/COOR+BOX) [BOUNDARY] type = PBC box_size_x = 68.25815 box_size_y = 80.24045 box_size_z = 66.58892 domain_x = 2 domain_y = 2 domain_z = 2 num_cells_x = 136 num_cells_y = 160 num_cells_z = 132 [SELECTION] group1 = ai:1-892 group2 = rnam:TIP3 and an:OH2 [SPANA_OPTION] buffer = 10 wrap = yes box_size = TRAJECTORY [RDF_OPTION] rmode = proximal solute = 1 solvent = 2 binsize = 0.25 range = 10 bulk_region = 15.0 recenter = 1