[gmx-users] Replica exchange probabilities beyond version 2016

Gül Zerze h.gul.ozer at gmail.com
Tue Apr 14 04:55:50 CEST 2020


Hi Pritam,

Did you check your energy fluctuations? Looks like you are using NH
thermostat, you might be suffering from the interaction of NH thermostat
oscillations with resonance modes in the system, which makes energy
fluctuations substantially larger (compared to fluctuations with NH in
older versions of Gromacs). As a result of amplified fluctuations, you have
higher exchange probability. See here for a previous relevant exchange:
https://redmine.gromacs.org/issues/2904 For my system, making tau_t=10
helped to recover the old behavior.

Best,
Gül
*--*
*Gül H. Zerze*
Postdoctoral Research Associate

Chemical and Biological Engineering
Princeton University
http://pablonet.princeton.edu/



On Mon, Apr 13, 2020 at 8:01 PM Pritam Ganguly <pganguly at chem.ucsb.edu>
wrote:

> Hello,
>
> I have noticed that the average exchange probability for a replica
> exchange run increases significantly if I use Gromacs versions 2018, as
> compared to versions 2016 or earlier. I have tested that by simulating
> the same system with versions 2016.4 and 2018.3 and the average
> probabilities I find are:
>
> Version 2016.4:
>
> Replica exchange statistics
> Repl  1666 attempts, 833 odd, 833 even
> Repl  average probabilities:
> Repl     0    1    2    3    4    5    6    7    8    9   10 11 12
> 13   14   15   16   17   18   19   20   21   22   23 24 25   26   27
> 28   29   30   31   32   33   34   35   36 37 38   39   40   41   42
> 43   44   45   46   47   48   49 50 51   52   53   54   55   56   57
> 58   59   60   61   62 63
> Repl      .29  .33  .29  .30  .28  .30  .30  .29  .29  .29  .30 .26
> .29  .30  .28  .29  .26  .26  .29  .28  .26  .27  .27  .27 .27  .28
> .26  .27  .26  .25  .27  .25  .26  .27  .25  .24  .25 .26  .27  .25
> .27  .26  .26  .26  .25  .26  .26  .24  .25  .27 .26  .27  .26  .25
> .27  .27  .28  .26  .27  .26  .27  .27  .28
>
>
> Version 2018.3:
>
> Replica exchange statistics
> Repl  1666 attempts, 833 odd, 833 even
> Repl  average probabilities:
> Repl     0    1    2    3    4    5    6    7    8    9   10 11 12
> 13   14   15   16   17   18   19   20   21   22   23 24 25   26   27
> 28   29   30   31   32   33   34   35   36 37 38   39   40   41   42
> 43   44   45   46   47   48   49 50 51   52   53   54   55   56   57
> 58   59   60   61   62 63
> Repl      .39  .39  .37  .42  .38  .38  .39  .36  .36  .38  .40 .39
> .40  .39  .37  .35  .40  .39  .35  .38  .40  .35  .38  .35 .36  .35
> .36  .36  .38  .36  .37  .36  .35  .36  .36  .37  .35 .35  .35  .37
> .37  .35  .37  .36  .37  .34  .36  .38  .38  .38 .35  .37  .36  .39
> .36  .40  .38  .38  .40  .36  .37  .36  .33
>
> These are test runs for 5 ns with exchange frequencies of 3 ps. I have
> used the same input parameters for both of these runs (mdp is attached).
> I actually observed this with many other systems and I was looking for
> any relevant gromacs release note for versions 2018 or beyond related to
> the replica exchange algorithm or the way the energies are calculated,
> but I could not figure it out.
>
> Regards,
>
> Pritam
>
> --
> Pritam Ganguly
> Department of Chemistry and Biochemistry
> University of California at Santa Barbara
> Santa Barbara, CA, USA 93106
>
> Office: PSBN 3606
> Phone: +1-805-893-2767
> http://people.chem.ucsb.edu/ganguly/pritam
>
> --
> Gromacs Users mailing list
>
> * Please search the archive at
> http://www.gromacs.org/Support/Mailing_Lists/GMX-Users_List before
> posting!
>
> * Can't post? Read http://www.gromacs.org/Support/Mailing_Lists
>
> * For (un)subscribe requests visit
> https://maillist.sys.kth.se/mailman/listinfo/gromacs.org_gmx-users or
> send a mail to gmx-users-request at gromacs.org.


More information about the gromacs.org_gmx-users mailing list