From: James Starlight (jmsstarlight_at_gmail.com)
Date: Tue Sep 17 2013 - 15:30:31 CDT
Also I've noticed that charm-gui inp files uses slightly differ simulation
setup ( in particular shorter cut-offs and lower relaxation times in
barostat) than in NAMD membrane protein tutorial.
>>> Namd works fine to simulate membrane proteins using standard
parameters, don't try to change all the simulation parameters to make it
work. You're wasting your time. It most probably doesn't work because your
input structure and/or psf files have problems.
2013/9/17 Niklaus Johner <niki.johner_at_gmail.com>
> 1) even protonated residues have integer charges, except if you screwed up
> when patching your residues, which is what I suspect happened. Did you get
> any warnings when generating your psf with psfgen, typically that he had to
> guess the position of heavy atoms, or even that he failed to set the
> position of some atoms? My guess is you did something wrong when
> protonating these residues. If they look separated from the residues they
> belong to when you visualize them, something is wrong (be sure to load both
> the psf and pdb files, so that the bonds represented in vmd are the actual
> bonds you have in your structure).
> 2) Are you loading a psf generated with psfgen or by charm-gui? Everything
> that is generated by charm-gui is usually fine, so as you seem new to all
> of this I suspect you introduced some problems later. What happens if you
> simulate the solvated system created by charm-gui, without any of your
> 3)If you can't even load your psf/pdb pair into vmd, you probably have
> something wrong in your files, no point in trying to simulate that, instead
> look for the error in your files. If it looks wrong, it's probably wrong!
> So protons floating around are not normal, and mean your patch wasn't
> applied properly.
> 4) Namd works fine to simulate membrane proteins using standard
> parameters, don't try to change all the simulation parameters to make it
> work. You're wasting your time. It most probably doesn't work because your
> input structure and/or psf files have problems.
> In conclusion find out what the problem in your pdb/psf files is! Good
> advice is to do this systematically to find out in which of your steps the
> problem is created. Also look at the output of psfgen, it is pretty helpful
> to identify problematic atoms. You can also set the output frequency to 1
> so that even when the simulation crashes rapidly you will have some
> trajectory to look at. Typically during minimization and/or MD you'll be
> able to see which atoms move a lot, which bonds stretch exagerately and so
> on, which will help you identify the error in your files.
> Niklaus Johner
> Weill Cornell Medical College
> Harel Weinstein Lab
> Department of Physiology and Biophysics
> 1300 York Avenue, Room D-501
> New York, NY 10065
> On Sep 17, 2013, at 1:12 PM, James Starlight wrote:
> 1) The possible source of error with non-integer charge could arise from
> protonation state set to some titrable residues (e.g I've chosen one asp
> and another glu to be protonated by means of psfgen script)
> but during visualization I've realized that two new added protons looks
> like separate atoms. Also during minimization of such system I've obtained
> some erros with PME so I suppose that such sustem have been prepared with
> errors initially)
> 2) Another problem with PSF file generated by psfgen. I've tried to load
> this psf with the membnrane.pdb (both files created by charm-gui) to vmd
> and obtained error
> psfgen) Created by CHARMM version 36 1
> psfgen) clearing structure, preserving topology and aliases
> psfgen) reading structure from psf file membrane.psf
> psfgen) Error processing bonds
> does it possible to make new PSF via psfgen for membrane.pdb (Assuming
> that my membrane consist of popc pope tip3p as well as ions) ?
> 2013/9/17 Aron Broom <broomsday_at_gmail.com>
>> if you have non-integer net charge for your system, that suggests there
>> may be more important general problems with your system topology. In
>> general, the forcefields tend to be setup in such a way as to give integer
>> or very close to integer values.
>> You may want to inspect all the partial charges and see that they both
>> make sense, and that each individual molecule in your system has an integer
>> charge (i.e. find the molecule that doesn't)
>> On Tue, Sep 17, 2013 at 1:33 AM, James Starlight <jmsstarlight_at_gmail.com>wrote:
>>> Hi Norman!
>>> I've already tried to increased
>>> langevinPistonPeriod 200.
>>> langevinPistonDecay 50.
>>> up to
>>> langevinPistonPeriod 400.
>>> langevinPistonDecay 100.
>>> but simulation have been crushed in any case
>>> by the way the possible source of error could be due to non integer
>>> charge that my system has.
>>> Initialy I had protein with ligand with total charge +3 and neitralized
>>> membrane produced by charm-gui (non charged lipids). After insertion of my
>>> protein into membrane and re-ionization I have obtained total charge 0.45.
>>> I have no idea how I could fix it yet :(
>>> 2013/9/16 Norman Geist <norman.geist_at_uni-greifswald.de>
>>>> Hi James,****
>>>> ** **
>>>> if you’re using a barostat, consider increasing the relax/decay times,
>>>> as fixed atoms lead to high forces due the coordinate rescaling which moves
>>>> the non-fixed into the fixed atoms. Otherwise check your minimization log
>>>> file for messages like “Moving xxx Atoms with bad contacts downhill” or
>>>> even worse “Giving up on xxx Atoms with bad contacts”. If so, you might
>>>> have some superimposed atoms in your structure. Also check if your
>>>> minimization was run for long enough by plotting the TOTAL energy and see
>>>> if it converged.****
>>>> ** **
>>>> Norman Geist.****
>>>> ** **
>>>> *Von:* owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] *Im
>>>> Auftrag von *James Starlight
>>>> *Gesendet:* Montag, 16. September 2013 14:08
>>>> *An:* namd-l_at_ks.uiuc.edu
>>>> *Betreff:* Fwd: namd-l: charm-gui membrane builder****
>>>> ** **
>>>> Recently I've forced with new problem during simulation of the membrane
>>>> Following Namd's tutorial I've built my system consisted of receptor
>>>> embedded in the popc bilayer surrounded water and ions.
>>>> After minimization I've launch first stage of equilibration (fixed all
>>>> atoms excepts of the lipid tales)****
>>>> this simulation have been crashed immediately with rattle errors.****
>>>> than I've decreased timestep down to 0.3 and increased cutoffs up to 25
>>>> (because namd told me than I should do it). Within this setup I can run my
>>>> simulation but on each step I obtain
>>>> ERROR: Margin is too small for 2 atoms during timestep 5001.
>>>> ERROR: Incorrect nonbonded forces and energies may be calculated!****
>>>> I have no any margin in my conf file so I have no idea how I can fix it
>>>> and what possible source of errors.****
>>>> Thanks for help****
>> Aron Broom M.Sc
>> PhD Student
>> Department of Chemistry
>> University of Waterloo
This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:21:41 CST