From: Jérôme Hénin (jerome.henin_at_ibpc.fr)
Date: Tue Feb 21 2017 - 09:51:20 CST
Hi Randy,
You seem to have many moving restraints working at the same time. They are
evaluated in parallel over SMP, in no particular order - that explains the
messy output you get. The real problem here is that some of these messages
do not specify which restraint they come from. I assume you want to compute
a total free energy for all of those, in which case you are really
interested in the sum of the free energy derivatives, and it doesn't really
matter which one is which - still, I'm sure it will be better to have each
one properly identified.
That's an easy fix, I've just done it on our source repository:
https://github.com/colvars/colvars
Best,
Jerome
On 21 February 2017 at 16:03, Randy J. Zauhar <r.zauhar_at_usciences.edu>
wrote:
> Hi Giacomo and fellow experts,
>
> I am running a thermodynamic cycle to evaluate a free energy of binding of
> a peptide to a target protein.
>
> One of the simulations is to release colvar constraints, and I found free
> energy change by integrating dA/dLambda values from the log file. The
> simulation ran without a hitch and the energy change was ‘reasonable’.
>
> Now two separate questions:
>
> 1) Next I want to convince myself of reversibility and run the the lambda
> schedule backwards to create the restraints. I use the endpoint of
> constraint annihilation as starting point. The simulation seems to run fine
> for a while, but died relatively early on with no error message. Assuming
> this was a fluke, I just ran it again - same issue. This time I restarted
> (it died at lambda = 0.098), and it got all the way to lambda = 0.77, but
> then the restart died with this message:
>
> ERROR: Constraint failure; simulation has become unstable.
> ERROR: Exiting prematurely; see error messages above.
>
> But I swear there IS NO error message that I can find ‘above', and nothing
> looks amiss in the colvars output.
>
> I am using NAMD 2.12b1 for Linux multicore, and have been running with 48
> threads like this:
>
> nohup namd +p 48 +isomalloc_sync complexPhosphoSTAT5RestrainCreateRestraint.RESTART.namd
> >& complexPhosphoSTAT5RestrainCreateRestraint.RESTART.log &
>
> 2) I have assumed that the dA/dLambda entries in the log file follow the
> same order as I specified in the colvars input file - but If I look more
> carefully in the log file, I see this sort of report:
>
> FINISHED WRITING RESTART VELOCITIES
> colvars: Lambda= 0.1 dA/dLambda= 0.0850971
> colvars: Lambda= 0.1 dA/dLambda= 10.3175
> colvars: Lambda= 0.1 dA/dLambda= 0.0294449
> colvars: Lambda= 0.1 dA/dLambda= 0.194364
> colvars: Lambda= 0.1 dA/dLambda= 39.9485
> colvars: Lambda= 0.1 dA/dLambda= 2.64137
> colvars: Restraint harmonic3, stage 2 : lambda = 0.11
> colvars: Lambda= 0.1 dA/dLambda= 0.494314
> colvars: Lambda= 0.1 dA/dLambda= 2.34284
> colvars: Lambda= 0.1 dA/dLambda= 3.04824
> colvars: Lambda= 0.1 dA/dLambda= 0.22976
> colvars: Restraint harmonic18, stage 2 : lambda = 0.11
> colvars: Restraint harmonic4, stage 2 : lambda = 0.11
> colvars: Setting force constant to 0.55
> colvars: Lambda= 0.1 dA/dLambda= 9.93387
> colvars: Lambda= 0.1 dA/dLambda= 3.78617
> colvars: Setting force constant to 0.55
> colvars: Restraint harmonic6, stage 2 : lambda = 0.11
> colvars: Setting force constant to 0.55
> colvars: Lambda= 0.1 dA/dLambda= 0.110793
> colvars: Lambda= 0.1 dA/dLambda= 0.0482745
> (snip) -------------------
>
> The reports for updating lambda seem to appear in random order - there is
> the sense that more than one thread is writing at the same time. Am I
> guaranteed that the derivatives are reported in a consistent order? Or, is
> there a better way to collect this information?
>
> THANKS in advance.
>
> Randy
>
>
> Randy J. Zauhar, PhD
>
> Prof. of Biochemistry
>
> Dept. of Chemistry & Biochemistry
> University of the Sciences in Philadelphia
> 600 S. 43rd Street
> Philadelphia, PA 19104
>
> Phone: (215)596-8691
> FAX: (215)596-8543
> E-mail: r.zauhar_at_usciences.edu<mailto:r.zauhar_at_usciences.edu>
>
>
>
> "The economy is still largely in ruins, thanks to the people who “drive
> the economy” by doing imaginary things on Wall Street, and there just isn’t
> much money left to spare for people who do actual things anymore."
>
> -- Tim Kreider, writing in the NY Times
>
>
>
This archive was generated by hypermail 2.1.6 : Mon Dec 31 2018 - 23:20:07 CST