[Chimera-users] Bug (?) with EnsembleCluster

Tue Jan 15 06:44:21 PST 2008

Got FMI, thanks. That seems a reasonable rationalization. Though, if I
understand, importing pdb files without saturating memory (may be importing one
"model", analyzing it, and then removing it from memory before importing next
"model", or any other better idea) is something to overcome in order to have
Chimera carrying out cluster analysis. Or I do not understand or cluster
analysis is of minor interest, both equally possible.

francesco

--- Eric Pettersen <pett at cgl.ucsf.edu> wrote:

> Just FYI, when you open a 5000 MODEL structure with Chimera's normal  
> File...Open, you get 5000 completely separate models (and will likely  
> exhaust memory unless its a small molecule).  Each model can be  
> colored individually, moved, mutated, closed, etc. individually.   
> When you open the same 5000 MODEL file with the MD Movie extension  
> you get _one_ model with 5000 sets of coordinates, which takes much  
> less memory.  On the other hand, each frame shares colors and so  
> forth with all the other frames, cannot be moved with respect to the  
> other frames, etc.  VMD undoubtedly uses the latter approach.
> 
> --Eric
> 
>                          Eric Pettersen
>                          UCSF Computer Graphics Lab
>                          http://www.cgl.ucsf.edu
> 
> On Jan 11, 2008, at 3:04 PM, Francesco Pietra wrote:
> 
> > Yes, 5365. What happened will be discovered. Up to the stage on
> > amber_score_everything, all was perfectly OK. Minimization and  
> > heating was also
> > OK. As to MD, the situation could not be clarified so far.
> >
> > In view of your comments below, I carried out a check with proven  
> > MD. I went to
> > Amber tutorial B3, took equil1.mdcrd.gz, decompressed, took the  
> > related prmtop,
> > and fed the two files to Chimera, last Daily Building. After all  
> > 5000 steps had
> > been acquired, saved pdb, all frames. Quit MD Movie. Try to open  
> > this pdb.
> > Well, with my modest desktop, Debian Linux i386, 1GB ram, this pdb  
> > could not be
> > opened: rapidly the memory was exhausted and to get control of the  
> > computer I
> > had to kill gnome. With vmd, this pdb developed rapidly with all  
> > 5000 frames.
> > Exactly as with the pdb from MD with my protein. This does not mean  
> > that my MDs
> > are OK. I'll look for what may be wrong there.
> >
> > Thanks
> > francesco
> >
> >
> >
> >
> > --- Eric Pettersen <pett at cgl.ucsf.edu> wrote:
> >
> >> 5365 atoms?  Given average atom counts in protonated protein residues
> >> that means your ligand came in close contact with ~340 residues over
> >> the course of the trajectory -- so it's wandering all over the
> >> place.  I can see why you want to use cluster analysis.  Oh well, I
> >> guess we're stuck until I have time to coordinate with Conrad and
> >> integrate his clustering code with MD Movie.
> >>
> >> --Eric
> >>
> >> On Jan 10, 2008, at 6:38 AM, Francesco Pietra wrote:
> >>
> >>> Eric: I followed previous route from combined mdcrd from ptraj.
> >>>
> >>> Removed remaining water.
> >>>
> >>> select ligand z<2.5
> >>>
> >>> as Chimera command; (this was the minimum, selecting 244 atoms, 118
> >>> of which
> >>> for the ligand; with z<2.0 the ligand only is selected)
> >>>
> >>> On playing the pdb (saved for selected atoms only) (with LOOP
> >>> deselected) the
> >>> atom selection increased, reaching 5365 atoms at the last (549)
> >>> frame. In
> >>> another run, with z<4.0 the course of atom selection was similar,
> >>> reaching 5993
> >>> atoms at frame 549.
> >>>
> >>> This huge number of atoms proved problematic for opening pdb. For
> >>> the z<2.5
> >>> case I left the computer on for 4h and half, with python occupying
> >>> 98% of the
> >>> available MEM (nearly 1GB). I had to kill gnome to recover command.
> >>>
> >>> May be with much more MEM it will work, though I can't check that  
> >>> now.
> >>>
> >>> I did not try to combine the trajectories with Chimera as I have no
> >>> indication
> >>> that those from ptraj are faulty. The pdb files for both z<2.5 and
> >>> z<4.0 open
> >>> in vmd.
> >>>
> >>> francesco
> >>>
> >>> --- Eric Pettersen <pett at cgl.ucsf.edu> wrote:
> >>>
> >>>> On Jan 9, 2008, at 3:32 PM, Francesco Pietra wrote:
> >>>>
> >>>>> In the "Define script.." I defined "select ligand z<10", then OK
> >>>>> and a BUG
> >>>>> window was presented. Attached is the reply log (bug). I tried  
> >>>>> with
> >>>>> z<4 with
> >>>>> the same result (actually, I don't remember if attached bug file
> >>>>> refers to z<10
> >>>>> or z<4).
> >>>>
> >>>> In the script-definition dialog you need to have "Interpret script
> >>>> as" set to "Chimera commands".  You had it set to "Python".
> >>>>
> >>>>> Still found difficulties in loading more than one mdcrd, though
> >>>>> this should be
> >>>>> my fault. I was not sure if the list of mdcrd is to be given at
> >>>>> beginning or if
> >>>>> the other mdcrd files after the first one are to be given
> >>>>> subsequently from the
> >>>>> menu. I tried with the "list" option but only the first mdcrd was
> >>>>> opened.
> >>>>
> >>>> When you run MD Movie and the dialog comes up where you specify the
> >>>> prmtop/trajectory files, you can use the "Add..." button to add as
> >>>> many trajectory files as needed.  I think this means you "give  
> >>>> it at
> >>>> the beginning" rather than "subsequently from the menu" in your
> >>>> terminology.
> >>>>
> >>>> You can also just list the files in a "metafile" for the command
> >>>> line.  For instance, this file works for me:
> >>>>
> >>>> amber
> >>>> leap.top
> >>>> md01.crd
> >>>> md02.crd
> >>>> md03.crd
> >>>> md04.crd
> >>>> md05.crd
> >>>> md06.crd
> >>>>
> >>>> Note that there _still_ seems to be something funky with Amber
> >>>> compressed trajectories, so don't use them for now but I should  
> >>>> have
> >>>> things fixed in a day or so with that.
> >>>>
> >>>> --Eric
> >>>>
> >>>>
> >>>
> >>>
> >>>
> >>>
> >>>
> >>> ____________________________________________________________________ 
> >>> __
> >>> ______________
> >>> Be a better friend, newshound, and
> >>> know-it-all with Yahoo! Mobile.  Try it now.  http://
> >>> mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ
> >>>
> >>
> >>
> >
> >
> >
> >        
> > ______________________________________________________________________ 
> > ______________
> > Looking for last minute shopping deals?
> > Find them fast with Yahoo! Search.  http://tools.search.yahoo.com/ 
> > newsearch/category.php?category=shopping
> 
> 

      ____________________________________________________________________________________
Be a better friend, newshound, and 
know-it-all with Yahoo! Mobile.  Try it now.  http://mobile.yahoo.com/;_ylt=Ahu06i62sR8HDtDypao8Wcj9tAcJ 