[Dock-fans] Fwd: Large ligand on Large grid Kill mpirun

Francesco Pietra chiendarret at yahoo.com
Sun Nov 18 00:54:10 PST 2007


There are 4GB ram (in 2GB slots Kingston ecc 400MHz) per processor (for a total
of 16GB ram) and Linux is set to let all that be used. Perhaps you may suggest
how to set the configuration better in order that all ram is used by DOCK.

With the slightly smaller ligands, where DOCK run correctly, I detected through
"top -i" a max of 24% memory used by one of the two processors in use. All four
processors were only used during amber rescore, and only for the initial half
reriod of the procedure. Then, the number dropped to two even for amber
rescore.
Thanks
francesco

--- Scott Brozell <sbrozell at scripps.edu> wrote:

> Hi,
> 
> The failure occurs in Orient::match_ligand.
> This is the focal point of much memory consumption.
> The simplest patch would be additional hardware memory.
> 
> Scott
> 
> On Fri, 16 Nov 2007, Francesco Pietra wrote:
> 
> > I forgot an important observation: docking procedures went on regularly
> even
> > for the largest (155 and 165 atoms) ligands when, on previous runs, the
> spheres
> > covered ca 1/4 of the protein.
> >
> > --- Francesco Pietra <chiendarret at yahoo.com> wrote:
> >
> > > Date: Fri, 16 Nov 2007 02:11:22 -0800 (PST)
> > > From: Francesco Pietra <chiendarret at yahoo.com>
> > > Subject: Large ligand on Large grid Kill mpirun
> > > To: dock-fans <dock-fans at docking.org>
> > >
> > > Following careful comparative runs I arrived at the conclusion that too
> large
> > > ligands on large grids kill mpirun.
> > >
> > > With grid files and selected_spheres.sph (10,536,750 grid points), for
> > > ligands
> > > of 118 atoms, or less, docking and amber rescore procedures went to
> > > completion
> > > OK. Selected spheres was centered (Magis' sphere_select) symmetrically in
> the
> > > protein for a 25A radius, whereby the spheres covered most of the
> protein.
> > >
> > > mpirun failure occurred with two ligands of much similar shape as above
> > > though
> > > larger (155 atoms and 165 atoms)
> > >
> > >
> > > How mpirun was killed shortly after launching the rigid score procedure,
> is
> > > shown by outputting the screen events following "mpirun -np 4 -i rigid.in
> -o
> > > rigid.out 2>&1 | tee screen.out":
> > >
> > > Initializing MPI Routines...
> > > Initializing MPI Routines...
> > > Initializing MPI Routines...
> > > Initializing MPI Routines...
> > > terminate called after throwing an instance of 'std::bad_alloc'
> > >   what():  St9bad_alloc
> > > [deb64:03725] *** Process received signal ***
> > > [deb64:03725] Signal: Aborted (6)
> > > [deb64:03725] Signal code:  (-6)
> > > [deb64:03725] [ 0] /lib/libpthread.so.0 [0x2b21e08e0410]
> > > [deb64:03725] [ 1] /lib/libc.so.6(gsignal+0x3b) [0x2b21e0a1807b]
> > > [deb64:03725] [ 2] /lib/libc.so.6(abort+0x10e) [0x2b21e0a1984e]
> > > [deb64:03725] [ 3]
> > >
> /usr/lib/libstdc++.so.6(_ZN9__gnu_cxx27__verbose_terminate_handlerEv+0x114)
> > > [0x2b21e0583424]
> > > [deb64:03725] [ 4] /usr/lib/libstdc++.so.6 [0x2b21e05815a6]
> > > [deb64:03725] [ 5] /usr/lib/libstdc++.so.6 [0x2b21e05815d3]
> > > [deb64:03725] [ 6] /usr/lib/libstdc++.so.6 [0x2b21e05816ba]
> > > [deb64:03725] [ 7] dock6.mpi(main+0) [0x42c180]
> > > [deb64:03725] [ 8] /usr/lib/libstdc++.so.6(_Znwm+0x34) [0x2b21e0581954]
> > > [deb64:03725] [ 9] /usr/lib/libstdc++.so.6(_Znam+0x9) [0x2b21e0581a49]
> > > [deb64:03725] [10] dock6.mpi(_ZN6Orient12match_ligandER7DOCKMol+0x375)
> > > [0x447a85]
> > > [deb64:03725] [11] dock6.mpi(main+0xaf5) [0x42cc75]
> > > [deb64:03725] [12] /lib/libc.so.6(__libc_start_main+0xda)
> [0x2b21e0a054ca]
> > > [deb64:03725] [13] dock6.mpi(__gxx_personality_v0+0xc2) [0x41b4ea]
> > > [deb64:03725] *** End of error message ***
> > > mpirun noticed that job rank 0 with PID 3724 on node deb64 exited on
> signal
> > > 15
> > > (Terminated).
> > > 3 additional processes aborted (not shown).
> > >
> > > File rigid.out showed correct reading of grid.nrg.
> > >
> > > _______
> > > The procedure failed also on serial run with
> > >
> > > dock6 -i rigid.in -o rigid.out 2>&1 screen_serial.out
> > >
> > > screen_serial.out:
> > >
> > > terminate called after throwing an instance of 'std::bad_alloc'
> > > waht(): St9bad_alloc
> > >
> > >
> > > i.e, as it was already clear from the above, not a problem of mpi.
> > >
> > > ______
> > >
> > > I understand to be largely out of the main stream of docking procedures
> (and
> > > probably scope). Nonetheless, I am interested in how these large ligands
> > > behave. Therefore, how could I manage to compare the above ligands of
> various
> > > size?. I am wondering about changing the grid space and see if docking
> with
> > > the
> > > smaller ligands changes much or not. If OK I could try the larger ligands
> > > with
> > > the new grid. Any better idea? I am not considering to work without a
> grid
> > > because I want to compare several molecules. So far I used defaults from
> > > tutorials, i.e. for grid.in:
> > >
> > > compute_grids                  yes
> > > grid_spacing                   0.3
> > > output_molecule                no
> > > contact_score                  no
> > > energy_score                   yes
> > > energy_cutoff_distance         9999
> > > atom_model                     a
> > > attractive_exponent            6
> > > repulsive_exponent             12
> > > distance_dielectric            yes
> > > dielectric_factor              4
> > > bump_filter                    yes
> > > bump_overlap                   0.75
> > > receptor_file    /home/francesco/dockwork/grid/myprotein.mol2
> > > box_file            /home/francesco/dockwork/grid/rec_box.pdb
> > > vdw_definition_file /usr/local/dock6/parameters/vdw_AMBER_parm99.defn
> > > score_grid_prefix              grid
> > >
> > >
> > > All that was carried out using A. Magis' sphgen_cpp and sphere_select.
> > >
> > > Thanks
> > >
> > > francesco pietra
> 



      ____________________________________________________________________________________
Be a better sports nut!  Let your teams follow you 
with Yahoo Mobile. Try it now.  http://mobile.yahoo.com/sports;_ylt=At9_qDKvtAbMuh1G1SQtBI7ntAcJ


More information about the Dock-fans mailing list