Re: Constraint failure in RATTLE algorithm

From: Aron Broom (broomsday_at_gmail.com)
Date: Sun Jun 22 2014 - 23:02:28 CDT

so it appears as though there are two spots where the problem arises, and I
can imagine it might be related to angle problems from the patching. Sadly
I have no experience with that, so I can't add anything on how to fix it.

What I might suggest, just so that you are even more confident it is the
patching, is to remove the waters (you'll need to make a new psf) and run
that and see if the same issue happens. It didn't look like there were any
weird VDW clashes or anything, but simplifying the problem to a vacuum
might be informative, plus it makes testing any fixes easier, since you can
simulate for longer times very quickly.

On Sat, Jun 21, 2014 at 7:43 PM, Tristan Croll <tristan.croll_at_qut.edu.au>
wrote:

> Let me see if I understand how you've put this together...
>
> You've got your special "linker" residue OL1 which is a 10-carbon chain
> with a peptide backbone on each end, and a methyl group in place of the
> alpha hydrogen. It's effectively symmetrical, but you've got one end
> defined as the "backbone" and inserted as resid 3038, and are patching the
> other end in between 3045 and 3047. So I take it...
>
> "patch DELB" deletes the existing amide bond between 3045 and 3047; "patch
> STCN" makes a new amide bond between 3045 and 3038; patch STNC makes the
> bond between 3038 and 3047... what does CMPS do?
>
> I'll let someone more knowledgeable than me comment on the behaviour
> you're seeing... but may I suggest an easier way to get to the same end
> goal? Why not chop your OL1 residue in half? Define it as simply a
> (C-alpha) methylated amino acid with a linear 5-carbon sidechain, and
> insert a copy at positions 3038 and 3046. Then it's one very simple patch
> to link the two copies together at their terminal carbons. The fewer (and
> simpler) patches you introduce, the less likely you are to introduce
> problems - and those problems you do introduce will be easier to
> troubleshoot.
>
>
> Cheers,
>
> Tristan
>
> -----Original Message-----
> From: owner-namd-l_at_ks.uiuc.edu [mailto:owner-namd-l_at_ks.uiuc.edu] On
> Behalf Of Douglas Houston
> Sent: Friday, 20 June 2014 11:32 PM
> To: Aron Broom
> Cc: Namd Mailing List
> Subject: Re: namd-l: Constraint failure in RATTLE algorithm
>
> Hi Aron,
>
> Thanks, this allows me to reproduce the problem. The following files
> should be enough for anyone to visualise the instability:
>
>
> https://drive.google.com/file/d/0B3IbJv_UNHlWN3p0bTR6NjFER2M/edit?usp=sharing
>
> https://drive.google.com/file/d/0B3IbJv_UNHlWYXYwYkZCTmkzc2M/edit?usp=sharing
>
> https://drive.google.com/file/d/0B3IbJv_UNHlWQmlfTjhoNjZUZWc/edit?usp=sharing
>
> Apart from the problem atoms flying around, I notice the whole system
> is "pulsing" - is this normal or indicative of a problem with e.g.
> pressure?
>
> cheres,
> Doug
>
>
> Quoting Aron Broom <broomsday_at_gmail.com> on Thu, 19 Jun 2014 11:37:59
> -0400:
>
> > You'd just need to save that frame's coordinates using VMD, and then use
> > those as an input for a "new" simulation. But of course don't do any
> > minimization.
> >
> >
> > On Thu, Jun 19, 2014 at 10:51 AM, Douglas Houston <
> DouglasR.Houston_at_ed.ac.uk
> >> wrote:
> >
> >> Hi Aron,
> >>
> >> I had considered starting from just before where the distortion sets in
> >> but wasn't sure how to do that?
> >>
> >> As I understand it the "firsttimestep" keyword doesn't specify the frame
> >> to start on, merely where to start the numbering. Is there another
> keyword
> >> I need to use to get the simulation to start from a particular frame?
> >>
> >> cheers,
> >> Doug
> >>
> >>
> >>
> >>
> >> Quoting Aron Broom <broomsday_at_gmail.com> on Thu, 19 Jun 2014 10:15:49
> >> -0400:
> >>
> >> You're right Normal, I did also mean DCDFreq.
> >>>
> >>> Douglas,
> >>>
> >>> What happens if you take your "penultimate frame" and restart from
> that,
> >>> does it fail again very quickly?
> >>>
> >>> I agree that if it's taking any more than a few ps to crash,
> minimization
> >>> likely isn't the problem. It sounds like there is a very sensitive
> >>> problem
> >>> that only appears under specific circumstances, maybe something with a
> >>> dihedral.
> >>>
> >>> ~Aron
> >>>
> >>>
> >>> On Thu, Jun 19, 2014 at 6:26 AM, Douglas Houston <
> >>> DouglasR.Houston_at_ed.ac.uk>
> >>> wrote:
> >>>
> >>> Hi all,
> >>>>
> >>>> Thanks for all the suggestions. A few more details:
> >>>>
> >>>> This is not particularly reproducible. The error may occur after 5ns
> or
> >>>> 50ns. I have been running 20ns simulations and usually they finish,
> but
> >>>> now
> >>>> I want to do a 200ns simulation.
> >>>>
> >>>> When I visualise the trajectory just before the failure I see that the
> >>>> residue the atom belongs to (3045) is heavily distorted in the final
> >>>> frame;
> >>>> it looks OK in the penultimate frame. I will rerun with smaller
> dcdfreq
> >>>> as
> >>>> Norman suggested. I will also increase the timestep from 0.5 to 2 fs
> as
> >>>> this will hopefully throw the error sooner (my current simulation
> failed
> >>>> after 3 days of running).
> >>>>
> >>>> I do 10,000 minimization steps to start. The fact that the system is
> >>>> stable for up to 50ns suggests to me that further initial minimization
> >>>> won't help ... ?
> >>>>
> >>>> I understand that I may well have have a problem with bad parameters,
> or
> >>>> a
> >>>> bad/incomplete topology. Residue 3045 has a number of patches applied
> to
> >>>> it.
> >>>>
> >>>> The following lists the psfgen commands I have been using to generate
> the
> >>>> topology:
> >>>>
> >>>> package require psfgen
> >>>> topology top_all36_prot_ole.rtf
> >>>> topology top_all36_lipid.rtf
> >>>> topology toppar_water_ions.str
> >>>> pdbalias residue HIS HSE
> >>>> pdbalias atom ILE CD1 CD
> >>>> segment A {pdb 5CstapleW2A.pdb
> >>>> first ACE
> >>>> last CT2}
> >>>> patch DELB A:3045 A:3047
> >>>> patch STCN A:3045 A:3038
> >>>> patch STNC A:3038 A:3047
> >>>> patch CMPS A:3045 A:3038 A:3047
> >>>> coordpdb 5CstapleW2A.pdb A
> >>>> regenerate angles dihedrals
> >>>> guesscoord
> >>>> writepdb 5CstapleW2A_psfgen.pdb
> >>>> writepsf 5CstapleW2A_psfgen.psf
> >>>>
> >>>>
> >>>>
> >>>>
> >>>>
> >>>> Quoting Axel Kohlmeyer <akohlmey_at_gmail.com> on Wed, 18 Jun 2014
> 14:23:05
> >>>> -0400:
> >>>>
> >>>> On Wed, Jun 18, 2014 at 1:29 PM, Douglas Houston
> >>>>
> >>>>> <DouglasR.Houston_at_ed.ac.uk> wrote:
> >>>>>
> >>>>> Hi all,
> >>>>>>
> >>>>>> I keep encountering the following fatal error:
> >>>>>>
> >>>>>> ERROR: Constraint failure in RATTLE algorithm for atom 189!
> >>>>>> ERROR: Constraint failure; simulation has become unstable.
> >>>>>> ERROR: Exiting prematurely; see error messages above.
> >>>>>>
> >>>>>> The atom itself varies. I have searched previous messages and tried
> >>>>>> the
> >>>>>> suggestions (smaller timestep, minimization overkill, etc.) to no
> >>>>>> avail.
> >>>>>>
> >>>>>> What else could I try to get my simulation to finish? I have
> attached
> >>>>>> my
> >>>>>>
> >>>>>>
> >>>>> you need to look at this from a different perspective. first you need
> >>>>> to find out the reason, not try to suppress it.
> >>>>>
> >>>>> how reproducible is this failure? how soon does this happen after the
> >>>>> start of your simulation. have you visualized your simulation around
> >>>>> the time of the failure and seen where exactly an atom experiences
> >>>>> (too) large forces. you may have a problem with bad parameters, or a
> >>>>> bad/incomplete topology (= .psf) file. or you are very very far away
> >>>>> from equilibrium and may need to do multiple iterations of
> >>>>> minimization and relaxation. and so on and so on. there are many
> ways,
> >>>>> but no simple general solution that works always.
> >>>>>
> >>>>> axel.
> >>>>>
> >>>>>
> >>>>> .conf file so you can see my system and the simulation parameters I
> am
> >>>>>
> >>>>>> specifying.
> >>>>>>
> >>>>>>
> >>>>>
> >>>>>
> >>>>> cheers,
> >>>>>
> >>>>>> Doug
> >>>>>>
> >>>>>> _____________________________________________________
> >>>>>> Dr. Douglas R. Houston
> >>>>>> Lecturer
> >>>>>> Institute of Structural and Molecular Biology
> >>>>>> Room 3.23, Michael Swann Building
> >>>>>> King's Buildings
> >>>>>> University of Edinburgh
> >>>>>> Edinburgh, EH9 3JR, UK
> >>>>>> Tel. 0131 650 7358
> >>>>>> http://tinyurl.com/douglasrhouston
> >>>>>>
> >>>>>>
> >>>>>> --
> >>>>>> The University of Edinburgh is a charitable body, registered in
> >>>>>> Scotland, with registration number SC005336.
> >>>>>>
> >>>>>>
> >>>>>>
> >>>>>
> >>>>> --
> >>>>> Dr. Axel Kohlmeyer akohlmey_at_gmail.com http://goo.gl/1wk0
> >>>>> College of Science & Technology, Temple University, Philadelphia PA,
> USA
> >>>>> International Centre for Theoretical Physics, Trieste. Italy.
> >>>>>
> >>>>>
> >>>>>
> >>>>>
> >>>>
> >>>>
> >>>> _____________________________________________________
> >>>> Dr. Douglas R. Houston
> >>>> Lecturer
> >>>> Institute of Structural and Molecular Biology
> >>>> Room 3.23, Michael Swann Building
> >>>> King's Buildings
> >>>> University of Edinburgh
> >>>> Edinburgh, EH9 3JR, UK
> >>>> Tel. 0131 650 7358
> >>>> http://tinyurl.com/douglasrhouston
> >>>>
> >>>> --
> >>>> The University of Edinburgh is a charitable body, registered in
> >>>> Scotland, with registration number SC005336.
> >>>>
> >>>>
> >>>>
> >>>>
> >>>
> >>> --
> >>> Aron Broom M.Sc
> >>> PhD Student
> >>> Department of Chemistry
> >>> University of Waterloo
> >>>
> >>>
> >>
> >>
> >>
> >> _____________________________________________________
> >> Dr. Douglas R. Houston
> >> Lecturer
> >> Institute of Structural and Molecular Biology
> >> Room 3.23, Michael Swann Building
> >> King's Buildings
> >> University of Edinburgh
> >> Edinburgh, EH9 3JR, UK
> >> Tel. 0131 650 7358
> >> http://tinyurl.com/douglasrhouston
> >>
> >> --
> >> The University of Edinburgh is a charitable body, registered in
> >> Scotland, with registration number SC005336.
> >>
> >>
> >>
> >
> >
> > --
> > Aron Broom M.Sc
> > PhD Student
> > Department of Chemistry
> > University of Waterloo
> >
>
>
>
>
> _____________________________________________________
> Dr. Douglas R. Houston
> Lecturer
> Institute of Structural and Molecular Biology
> Room 3.23, Michael Swann Building
> King's Buildings
> University of Edinburgh
> Edinburgh, EH9 3JR, UK
> Tel. 0131 650 7358
> http://tinyurl.com/douglasrhouston
>
>
> --
> The University of Edinburgh is a charitable body, registered in
> Scotland, with registration number SC005336.
>
>
>

-- 
Aron Broom M.Sc
PhD Student
Department of Chemistry
University of Waterloo

This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:20:55 CST