Re: Metadynamics problems

From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Tue May 25 2021 - 09:29:12 CDT

Sorry, I left the reply in draft accidentally.

I would try not to exceed ~0.6 kcal/mol (thermal energy at 300 K) added
every correlation time of the variables, and ideally stay well below that
rate.

Giacomo

On Mon, May 17, 2021 at 12:51 PM Alessandro Ruda <alessandro.ruda_at_su.se>
wrote:

> Thanks Giacomo for the prompt reply,
>
>
> Restarting the run didn't really work since it stopped again after 20
> minutes (after running for 11 h though). Doesn't look stable at all.
>
>
> The correlation time for the torsions I am looking at is around 100 ps.
>
> What value would you suggest to be the optimal one in such case?
>
>
> All the best,
>
> Alex
> ------------------------------
> *From:* Giacomo Fiorin <giacomo.fiorin_at_gmail.com>
> *Sent:* 14 May 2021 17:43:08
> *To:* NAMD list; Alessandro Ruda
> *Subject:* Re: namd-l: Metadynamics problems
>
> Hello Alex, if the instability is sporadic you can safely restart and
> continue running.
>
> The instability, however, could be due to the relatively high rate at
> which you add the Gaussian bias, 1 kcal/mol per 1000 steps. Not knowing
> the specifics of the system, it would help you to figure out what are the
> correlation times of the two dihedral angles from an equilibrium MD
> trajectory, which you should already have. The variables may simply not
> take up a bias that fast.
>
> Lastly, you're using NAMD 2.12, which is more than 4 years old.
> While it's true that the error is raised from a large force coming from
> Colvars, later versions like 2.14 may handle it better, or at least clarify
> more clearly the precise step where things go wrong. For Colvars
> specifically, there are a few advantages for using metadynamics
> specifically:
> https://urldefense.com/v3/__https://colvars.github.io/namd-2.14/colvars-refman-namd/colvars-refman-namd.html*sec:colvars_config_changes__;Iw!!DZ3fjg!qOhlt2m9TYK7BRGsQiGuJiq3akA1iqhn-By0VCKkmL4dndq0_mT4Vui3UsJXPHGwmA$
>
> Giacomo
>
> On Fri, May 14, 2021 at 5:50 AM Alessandro Ruda <alessandro.ruda_at_su.se>
> wrote:
>
>> Dear *NAMD comunity*,
>>
>>
>> I am running metadynamics on a set of two dihedral angles. I've tried
>> default metadynamics and well tempered. For both of them after a few
>> hours the run stops with errors like the following:
>>
>>
>> *namd2:207125 terminated with signal 11 at PC=fc9213 SP=7ffe44263fa0.
>> Backtrace:*
>>
>> */software/sse/manual/namd/2.12/i18/nsc1/namd2(_ZN12PmeRealSpace12fill_chargesEPPfS1_RiS2_PcS3_P11PmeParticle+0x2d43)[0xfc9213]*
>>
>> */software/sse/manual/namd/2.12/i18/nsc1/namd2(_ZN12PmeRealSpace12fill_chargesEPPfS1_RiS2_PcS3_P11PmeParticle+0x2d43)[0xfc9213]*
>>
>> and the log file states:
>>
>> *mpprun info: Job terminated with error*
>>
>> Normal molecular dynamics works well so it must be related to
>> metadynamics. This is the colvar file that i used:
>>
>> --------------------------------------------------------------------------------------------------------------------------------------
>>
>> *colvarsTrajAppend off*
>>
>> *colvar {*
>> * name Psi*
>> * width 2.5*
>> * dihedral {*
>> * group1 {*
>> * atomnumbers { 32 }*
>> * }*
>> * group2 {*
>> * atomnumbers { 22 }*
>> * }*
>> * group3 {*
>> * atomnumbers { 20 }*
>> * }*
>> * group4 {*
>> * atomnumbers { 21 }*
>> * }*
>> * }*
>> * lowerBoundary -180*
>> * upperBoundary 180*
>>
>> *}*
>>
>> *colvar {*
>> * name Phi*
>> * width 2.5*
>> * dihedral {*
>> * group1 {*
>> * atomnumbers { 33 }*
>> * }*
>> * group2 {*
>> * atomnumbers { 32 }*
>> * }*
>> * group3 {*
>> * atomnumbers { 22 }*
>> * }*
>> * group4 {*
>> * atomnumbers { 20 }*
>> * }*
>> * }*
>> * lowerBoundary -180*
>> * upperBoundary 180*
>> *}*
>>
>> *metadynamics {*
>> *wellTempered on*
>> *biasTemperature 1490*
>> *name metaPhiPsi*
>> *colvars Phi Psi*
>> *hillWeight 0.01*
>> *hillwidth 1.0*
>> *saveFreeEnergyFile no*
>> *dumpFreeEnergyFile yes #When useGrids and this option are on, the PMF
>> is written every colvarsRestartFrequency steps to the file outputName.pmf*
>> *writeHillsTrajectory on*
>> *newHillFrequency 100 *
>>
>> ---------------------------------------------------------------------------------------------------------------------------------------
>>
>> I have tried with different hillWeight values and the problem stays.
>> Can I do something or should I restart from the last step and continue
>> running and, in doing so,
>> how can I keep the old colvars data so that I restart from the last
>> points collected?
>>
>> Thank you in advance,
>>
>> All the best,
>> Alex
>>
>>
>>
> All the best,
> Alex
>
>
>

This archive was generated by hypermail 2.1.6 : Fri Dec 31 2021 - 23:17:11 CST