Must add that alternatively to "average structure" I tried a cluster analysis with MMTSB. The program refused to work on my trajectories for unclear reasons: the program is installed correctly and the input is correct (as verified by the program author). Therefore, I am eagerly waiting for Chimera being able to carry out cluster analysis (if I understood correctly your message to this regards) francesco --- Francesco Pietra <chiendarret@yahoo.com> wrote:
Date: Wed, 2 Jan 2008 23:04:40 -0800 (PST) From: Francesco Pietra <chiendarret@yahoo.com> Subject: Re: [Chimera-users] Slow dealing with pdb files To: Eric Pettersen <pett@cgl.ucsf.edu> CC: chimera <chimera-users@cgl.ucsf.edu>
Eric: Thanks a lot for this lesson, surely useful to many other guys, too.
Whether looking at an "average structure" or some other output structure was a problem I got no clarification from the Amber mailing list. Unfortunately, Pr Brozell is not active in this period on Dockfans to ask him. My question was, is the "average structure" representative of the interactions occurring between the protein and the ligand as resulting from the MD carried out? Although DOCK6.1 does that with amber score, it is only for implicit medium surrounding the complex. Ideally, I would have liked to estimate the free energy of interaction in the presence of explicit surrounding medium but Amber does not appear to have a way to that for a complex in a membrane (or anyway for a non-standard ligand). Therefore, what I am relying on, is the distance between the protein residues and the ligand. The closest the protein residues are to the ligand, the more they are considered to be relevant.
At any event, given the problem Chimera has encountered with an average structure, do you believe that mapping the protein environment around the ligand with Chimera's "zone" is correct? From your "lesson" I understand YES.
I would appreciate any comment or suggestion about that.
francesco
--- Eric Pettersen <pett@cgl.ucsf.edu> wrote:
On Jan 1, 2008, at 1:44 PM, Francesco Pietra wrote:
I am dealing with the average structure (a protein complex embedded in a POCP membrane and water solvated) derived with Amber's ptraj from a 1.5 ns MD.
Opening this pdb file in 1.2470 Chimera has become extremely slow. The file is 6.4MB. First, below the screen it is warned "Ignored bad PDB record found on line #", for lines from 1 to 114154. This may take some 10 minutes.
These are for the water ATOM records where the atom serial number and/ or residue number were "****" (what FORTRAN inserts when a number won't fit inside a field width).
After that, the warning message changes to "Computed secondary structure assignments (see reply log)" which lasts for longer than 1 hour and 20 minutes. During this time, "top" command shows that python is using 12% MEM and 99% CPU.
Due to the fact that this is an "average" structure, Chimera's estimation of the connectivity is bad for many parts of the structure -- particularly the POP residues in the membrane. This creates a rat's nest of intra-residue connectivity which the ring-finding algorithm (designed for "reasonable" structures) takes a long time to operate on. Normally Chimera wouldn't run ring-finding as a structure opens, but due some interesting naming of hydrogens in the POP residues (e.g. RH16) it assigns some of the hydrogens to be other elements (e.g. rhodium, as per PDB atom naming rules). Since rhodium is a metal, it wants to depict it as a sphere, which means it needs to know the radius, which in turn depends on the atom type, which needs to find rings...
Then, the graphics appears, with the membrane-protein-complex not centered in the water box.
This is due to the "****" waters being ignored.
I could then carry out rapid mapping of the protein residues around the single-residue ligand (select protein & :ligandname z<#), which was what I wanted to do.
If you only care about the protein and ligand in your analysis, you should just edit your file to strip the waters and lipids. When I did this with the file you sent it only took moments to open.
--Eric
Eric Pettersen UCSF Computer Graphics Lab http://www.cgl.ucsf.edu
____________________________________________________________________________________
Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs
____________________________________________________________________________________ Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs
Hi Francesco, The daily builds from the last couple of weeks include an EnsembleCluster tool (under Tools... MD/Ensemble analysis) that you can use for trajectories in a single PDB file. If you had a different format, how to convert to single PDB is described here: http://www.cgl.ucsf.edu/pipermail/chimera-users/2007-December/ 002142.html I would recommend using ptraj or some similar program to sample your trajectory at time intervals rather than using every single step, if you haven't done that already. Structures from very close time steps are highly correlated with each other and just increase your file size without increasing your information content. Also, as Eric mentioned, you could strip waters or lipid molecules if they are not involved in your analyses. Chimera daily builds can be downloaded here: http://www.cgl.ucsf.edu/chimera/alpha-downloads.html Documentation for EnsembleCluster (hasn't made it yet to the "official" web site): http://www.cgl.ucsf.edu/home/meng/docs/ContributedSoftware/ ensemblecluster/ensemblecluster.html This tool clusters the structures and identifies cluster representatives. My opinion is that these are much more meaningful than an "average structure" from a trajectory. Best, Elaine ----- Elaine C. Meng, Ph.D. meng@cgl.ucsf.edu UCSF Computer Graphics Lab and Babbitt Lab Department of Pharmaceutical Chemistry University of California, San Francisco http://www.cgl.ucsf.edu/home/meng/index.html On Jan 3, 2008, at 7:18 AM, Francesco Pietra wrote:
Must add that alternatively to "average structure" I tried a cluster analysis with MMTSB. The program refused to work on my trajectories for unclear reasons: the program is installed correctly and the input is correct (as verified by the program author). Therefore, I am eagerly waiting for Chimera being able to carry out cluster analysis (if I understood correctly your message to this regards) francesco
Elaine: Please see below. Thans francesco --- Elaine Meng <meng@cgl.ucsf.edu> wrote:
Hi Francesco, The daily builds from the last couple of weeks include an EnsembleCluster tool (under Tools... MD/Ensemble analysis) that you can use for trajectories in a single PDB file. If you had a different format, how to convert to single PDB is described here: http://www.cgl.ucsf.edu/pipermail/chimera-users/2007-December/ 002142.html
My browser tells that this html does not exist, and I was unable to catch it from the last December list. At any event, what I can have (or I am able to get) from ptraj is a series of pdb files for each step. Is the pipermail mentioned above able to combine the pdb files into a single pdb file?
I would recommend using ptraj or some similar program to sample your trajectory at time intervals rather than using every single step, if you haven't done that already.
For the case in point, during MD the coordinates were written every 1000 steps, so that I do not restrict ptraj further to this concern.
Structures from very close time steps are highly correlated with each other and just increase your file size without increasing your information content. Also, as Eric mentioned, you could strip waters or lipid molecules if they are not involved in your analyses.
Chimera daily builds can be downloaded here: http://www.cgl.ucsf.edu/chimera/alpha-downloads.html
Documentation for EnsembleCluster (hasn't made it yet to the "official" web site): http://www.cgl.ucsf.edu/home/meng/docs/ContributedSoftware/ ensemblecluster/ensemblecluster.html
This tool clusters the structures and identifies cluster representatives. My opinion is that these are much more meaningful than an "average structure" from a trajectory. Best, Elaine ----- Elaine C. Meng, Ph.D. meng@cgl.ucsf.edu UCSF Computer Graphics Lab and Babbitt Lab Department of Pharmaceutical Chemistry University of California, San Francisco http://www.cgl.ucsf.edu/home/meng/index.html
On Jan 3, 2008, at 7:18 AM, Francesco Pietra wrote:
Must add that alternatively to "average structure" I tried a cluster analysis with MMTSB. The program refused to work on my trajectories for unclear reasons: the program is installed correctly and the input is correct (as verified by the program author). Therefore, I am eagerly waiting for Chimera being able to carry out cluster analysis (if I understood correctly your message to this regards) francesco
____________________________________________________________________________________ Never miss a thing. Make Yahoo your home page. http://www.yahoo.com/r/hs
participants (2)
-
Elaine Meng -
Francesco Pietra