Re: colvars with replica exchange molecular dynamics

From: Francesco Pietra (chiendarret_at_gmail.com)
Date: Fri Jan 31 2014 - 01:38:21 CST

Hi Giacomo:

> If I understood correctly your email, some restart files are not correctly
> written before an exchange between replicas

occurs.
>

In the presence of colvars, a xxx.colvars.state file is written under each
replica folder for job0. However, a xxx.restart.colvars.state is written
nowhere.

If I restart a t-remd from the previous one, linking between the two occurs
correctly except for colvars. The restarted t-remd writes again
xxx.colvars.state under each replica folder for job1. However (as far as I
can understand) the state of colvars at the end of job0 is not taken into
consideration.

I tried to remedy by modifying replica.namd for job1, imitating the
procedure for umbrella sampling (although I have no experience with
umbrella sampling). My procedure was incorrect, as I described.

I assume that "replicaBarrier" is there and that T-REMD without a colvars
> configuration happens correctly, right?

Unfortunately I don't understand "replicaBarrier". As said above, T-REMD
runs in any case, colvars present or not. No problem in linking a new
T-REMD to a previous one, except for colvars.

What I am not sure about is whether colvars for job1 start from what the
colvars situation at the end of job0, the absence of
xxx.restart.colvars.state nothwithstanding.

>From your mail I understand that I am unable to describe the phenomenon
adequately. *I could provide a short T-REMD example with job0 and job1 and
colvars to you privately by sending files to your server if you give me
access*. Unfortunately the server at the Accademia Lucchese, where I am, is
inadequate to the purpose, while I have no command on servers of CINECA,
where I carry out the T-REMD runs.

Thanks
francesco pietra

On Fri, Jan 31, 2014 at 12:32 AM, Giacomo Fiorin
<giacomo.fiorin_at_gmail.com>wrote:

> If I understood correctly your email, some restart files are not correctly
> written before an exchange between replicas occurs. I assume that
> "replicaBarrier" is there and that T-REMD without a colvars configuration
> happens correctly, right?
>
>
> On Thu, Jan 30, 2014 at 5:38 AM, Francesco Pietra <chiendarret_at_gmail.com>wrote:
>
>> Hi Giacomo:
>> I forgot to mention that I was using namd2.10. However, in the recent
>> past, same issue with namd2.9 on the same BlueGene/Q
>> francesco pietra
>> ---------- Forwarded message ----------
>> From: Francesco Pietra <chiendarret_at_gmail.com>
>> Date: Wed, Jan 29, 2014 at 5:40 PM
>> Subject: Fwd: namd-l: colvars with replica exchange molecular dynamics
>> To: Giacomo Fiorin <giacomo.fiorin_at_gmail.com>, NAMD <namd-l_at_ks.uiuc.edu>
>>
>>
>> Hi Giacomo:
>> Back to BluegeneQ for T-REMD with colvars. I carried out other debug runs
>> with the same system of peptide under GB, non periodic, 32 replicas, temp
>> interval 300-600K.
>>
>> job0 (with standard replica.namd file) was carried out correctly, rmsd
>> colvars
>> ................................
>> rmsd {
>> atoms {
>> # add all the C-alphas within residues 41 to 48 of segment "P1"
>> psfSegID P1
>> atomNameResidueRange CA 41-48
>> }
>> refPositionsFile npt-01_autopsf.pdb }
>> }
>>
>>
>> # apply harmonic potential to the "rmsd" variable
>> # with center on zero
>> # colvars and centers should both be lists, appearing only once in the
>> block
>>
>> harmonic {
>>
>> name P2i_harmonic
>>
>> colvars { colvar1 }
>>
>> centers 0.0
>>
>> forceConstant 100.0
>> }
>> ....................................
>>
>> being transferred into each replica for restarting, such as for replica 0:
>>
>> fold_P2i.job0.0.colvars.state
>>
>> The folder /output contains the file "fold_P2ijob0.restart.100.tcl",
>> which reads:
>>
>> ............................
>> set i_job 1
>> set i_run 100
>> set i_step 10000
>> set restart_root output/%s/fold_P2i.job0.restart100
>> ........................
>> i.e., for restarting with job1, but without anything about colvars.
>> xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
>>
>> On restarting with job1, I added colvarsInput to replica.namd, in analogy
>> to indications for umbrella sampling, as follows;
>>
>> ......................
>>
>> if {$i_run} { #restart
>> bincoordinates $restart_root.$replica_id.coor
>> binvelocities $restart_root.$replica_id.vel
>> extendedSystem $restart_root.$replica_id.xsc
>> colvarsInput $restart_root.$replica_id.colvars.state
>> } else {
>> temperature $NEWTEMP
>> }
>> .............................
>>
>>
>> The error is now clear, in the sense that it occurs on all replicas, for
>> example replica 17 log file ends with:
>> .................................................
>> colvars: Initializing a new "harmonic" instance.
>> colvars: # name = P2i_harmonic
>> colvars: # colvars = { colvar1 }
>> colvars: # outputEnergy = off [default]
>> colvars: # forceConstant = 100
>> colvars: # centers = { 0 }
>> colvars: # targetCenters = { 0 } [default]
>> colvars: # targetForceConstant = 0 [default]
>> colvars: # outputCenters = off [default]
>> colvars: # outputAccumulatedWork = off [default]
>> colvars:
>> ----------------------------------------------------------------------
>>
>> colvars: Collective variables biases initialized, 1 in total.
>> colvars:
>> ----------------------------------------------------------------------
>> colvars: Error: in opening restart file
>> "output/17/fold_P2i.job0.restart100.17.colvars.state".
>> colvars: If this error message is unclear, try recompiling with
>> -DCOLVARS_DEBUG.
>>
>> FATAL ERROR: Error in the collective variables module: exiting.
>>
>> .................................................
>>
>> The same for other replicas, with their specific number in place of 17.
>> certain replicas ended at "colvars: Collective variables biases
>> initialized, 1 in total." just before the message of error. Showing the
>> error or not occurs randomly.
>>
>>
>> The little I can understand is that the trick I used does not pass
>> correctly the information about colvarsInput to the system. Standard
>> replica.namd does not work at all for restarting a T-REMD with colvars.
>>
>> Ask me please any other information about what I did, or suggest what I
>> could try in addition.
>>
>> Thanks
>> francesco pietra
>>
>> ---------- Forwarded message ----------
>> From: Francesco Pietra <chiendarret_at_gmail.com>
>> Date: Tue, Jan 14, 2014 at 9:25 AM
>> Subject: Re: namd-l: colvars with replica exchange molecular dynamics
>> To: Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
>> Cc: NAMD <namd-l_at_ks.uiuc.edu>
>>
>>
>> Giacomo:
>> Thanks so much for taking my post into consideration. That error message
>> was from a single t-remd with modified replica.namd. I need to run some
>> debugs in order to answer your question. That will take time if I move my
>> project from a BluGene with namd2.10 to a PLX machine with namd-cuda. I
>> read your answer as implying that our modification of replica.namd as for
>> umbrella sampling (suggested to me at CINECA) might be sound.
>>
>> francesco
>>
>>
>> On Mon, Jan 13, 2014 at 10:17 PM, Giacomo Fiorin <
>> giacomo.fiorin_at_gmail.com> wrote:
>>
>>> Francesco, is this a reproducible error, i.e. is it always replica 23
>>> which gives you the error message?
>>>
>>>
>>> On Mon, Jan 13, 2014 at 3:29 PM, Francesco Pietra <chiendarret_at_gmail.com
>>> > wrote:
>>>
>>>> Some time ago I asked much this same question which must have gone lost
>>>> or anyway raised no interest. Nonetheless, the issue is at the heart of
>>>> carrying out t-remd while restraining a portion of the molecule with
>>>> colvars.
>>>>
>>>> Setting colvars in namd config file for remd
>>>>
>>>> # constraints
>>>> colvars on
>>>> colvarsConfig ./P2i_rmsd-colvars.in
>>>>
>>>> gives correct
>>>> fold_P2i.job0.0.colvars.state
>>>> fold_P2i.job0.0.colvars.traj
>>>>
>>>> in all replicas and a restart file in /output folder. This restart file
>>>> does not provide any information about colvars. My question is where
>>>> information about colvars is provided. I can't see any.
>>>>
>>>> For umbrella sampling it is advised to add a line to the guiding file
>>>>
>>>> colvarsInput $restart_root.$replica_id.colvars.state
>>>>
>>>> to elicit information about colvars. On this basis, I added the same
>>>> line to replica.namd, as shown below:
>>>>
>>>>
>>>> if {$i_run} { #restart
>>>> bincoordinates $restart_root.$replica_id.coor
>>>> binvelocities $restart_root.$replica_id.vel
>>>> extendedSystem $restart_root.$replica_id.xsc
>>>> colvarsInput $restart_root.$replica_id.colvars.state
>>>> } else {
>>>> temperature $NEWTEMP
>>>> }
>>>>
>>>>
>>>> It did hot help; the .err file told that:
>>>>
>>>> REPLICA 23 FATAL ERROR: Error in the collective variables module:
>>>> exiting.
>>>>
>>>> The replica 23 log file, tells that:
>>>>
>>>> colvars: Collective variables biases initialized, 1 in
>>>> total.fold_P2i.job0.
>>>> restart500.23.colvars.state
>>>> colvars:
>>>> ----------------------------------------------------------------------
>>>> colvars: Error: in opening restart file
>>>> "output/23/fold_P2i.job0.restart500.23.colvars.state".
>>>>
>>>>
>>>>
>>>> The conclusion is that a t-remd with colvars cannot be properly
>>>> restarted as far as I can understand. Probably most t-remd simulations are
>>>> carried out without colvars, but there are situations where restraints on a
>>>> portion of the molecule are needed.
>>>>
>>>> Thanks for your attention
>>>>
>>>> francesco pietra
>>>>
>>>
>>>
>>
>>
>>
>

This archive was generated by hypermail 2.1.6 : Wed Dec 31 2014 - 23:22:04 CST