[gmx-users] REMD exchange probabilities
Mark Abraham
mark.j.abraham at gmail.com
Mon Mar 9 18:42:45 CET 2015
On Sun, Mar 8, 2015 at 7:25 PM, Neha Gandhi <n.gandhiau at gmail.com> wrote:
> Dear list,
>
> Using an exchange probability of 0.25 and temperature range 293-370 K, I
> calculated number of replicas using the server. However, when I did first
> run and tried exchanging replicas every 500 steps (1 ps), I don't think the
> exchange probabilities make sense in particular replicas 15 and 16. Replica
> 15 has a low exchange ratio of 0.12 while replica 16 has a high exchange
> ratio of 0.55.
>
This can be real. See literature from Nadler and Hansmann. If there's some
temperature-dependent change in the range of available configurations (e.g.
a phase transition), then you can have configuration(s) whose energy is
such that they are accessible at one temperature and not at adjacent
temperatures. Such replicas won't cross that temperature barrier until they
have found a region of phase space that permits it. Such an ergodicity
bottleneck suggests adding other replicas around that temperature, because
you need flow over replica space to achieve the desired enhanced sampling.
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
> Repl .28 .28 .28 .28 .29 .28 .29 .29 .28 .29 .28 .28 .29
> .29 .29 .12 .55 .29 .29 .30 .30 .29 .29 .26 .32 .31 .30 .30
> .30 .30 .30 .31 .31 .31 .31 .31 .31 .31 .31 .31 .31 .31 .32
> .32 .32 .32 .33
> Repl number of exchanges:
> 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
> Repl 2901 2954 2873 3017 3038 2910 3009 2993 2934 3002 2981 2999 2927
> 3038 3059 1229 5757 3056 3100 3136 3054 3053 3109 2743 3333 3166 3097 3185
> 3161 3189 3133 3226 3261 3242 3229 3205 3249 3227 3221 3222 3326 3303 3309
> 3320 3373 3346 3474
> Repl average number of exchanges:
> 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
> Repl .28 .28 .27 .29 .29 .28 .29 .29 .28 .29 .29 .29 .28
> .29 .29 .12 .55 .29 .30 .30 .29 .29 .30 .26 .32 .30 .30 .30
> .30 .31 .30 .31 .31 .31 .31 .31 .31 .31 .31 .31 .32 .32 .32
> .32 .32 .32 .33
>
>
> Below are the temperatures I have used. How do I manually edit temperatures
> to get average exchange probabilities between 0.2-0.3?
>
The same way you set up the original set of temperatures - make an .mdp
that has a temperature you want, equilibrate, and then insert it into the
set of replicas before a new run.
Your existing set of temperatures has one spacing of exactly one degree,
and the rest seem to be exponential, so that looks funny.
ref_t = 293 293 ; reference temperature, one for each
> group, in K
> ref_t = 294.51 294.51 ; reference temperature, one for each
> group, in K
> ref_t = 296.03 296.03 ; reference temperature, one for each
> group, in K
> ref_t = 297.56 297.56 ; reference temperature, one for each
> group, in K
> ref_t = 299.09 299.09 ; reference temperature, one for each
> group, in K
> ref_t = 300.63 300.63 ; reference temperature, one for each
> group, in K
> ref_t = 302.18 302.18 ; reference temperature, one for each
> group, in K
> ref_t = 303.73 303.73 ; reference temperature, one for each
> group, in K
> ref_t = 305.29 305.29 ; reference temperature, one for each
> group, in K
> ref_t = 306.86 306.86 ; reference temperature, one for each
> group, in K
> ref_t = 308.43 308.43 ; reference temperature, one for each
> group, in K
> ref_t = 310.01 310.01 ; reference temperature, one for each
> group, in K
> ref_t = 311.60 311.60 ; reference temperature, one for each
> group, in K
> ref_t = 313.19 313.19 ; reference temperature, one for each
> group, in K
> ref_t = 314.79 314.79 ; reference temperature, one for each
> group, in K
> ref_t = 316.40 316.40 ; reference temperature, one for each
> group, in K
> ref_t = 318.63 318.63 ; reference temperature, one for each
> group, in K
> ref_t = 319.63 319.63 ; reference temperature, one for each
> group, in K
> ref_t = 321.26 321.26 ; reference temperature, one for each
> group, in K
> ref_t = 322.89 322.89 ; reference temperature, one for each
> group, in K
> ref_t = 324.52 324.52 ; reference temperature, one for each
> group, in K
> ref_t = 326.17 326.17 ; reference temperature, one for each
> group, in K
> ref_t = 327.82 327.82 ; reference temperature, one for each
> group, in K
> ref_t = 329.49 329.49 ; reference temperature, one for each
> group, in K
> ref_t = 331.26 331.26 ; reference temperature, one for each
> group, in K
> ref_t = 332.86 332.86 ; reference temperature, one for each
> group, in K
> ref_t = 334.51 334.51 ; reference temperature, one for each
> group, in K
> ref_t = 336.20 336.20 ; reference temperature, one for each
> group, in K
> ref_t = 337.90 337.90 ; reference temperature, one for each
> group, in K
> ref_t = 339.61 339.61 ; reference temperature, one for each
> group, in K
> ref_t = 341.32 341.32 ; reference temperature, one for each
> group, in K
> ref_t = 343.04 343.04 ; reference temperature, one for each
> group, in K
> ref_t = 344.76 344.76 ; reference temperature, one for each
> group, in K
> ref_t = 346.49 346.49 ; reference temperature, one for each
> group, in K
> ref_t = 348.23 348.23 ; reference temperature, one for each
> group, in K
> ref_t = 349.98 349.98 ; reference temperature, one for each
> group, in K
> ref_t = 351.74 351.74 ; reference temperature, one for each
> group, in K
> ref_t = 353.50 353.50 ; reference temperature, one for each
> group, in K
> ref_t = 355.27 355.27 ; reference temperature, one for each
> group, in K
> ref_t = 357.04 357.04 ; reference temperature, one for each
> group, in K
> ref_t = 358.83 358.83 ; reference temperature, one for each
> group, in K
> ref_t = 360.62 360.62 ; reference temperature, one for each
> group, in K
> ref_t = 362.42 362.42 ; reference temperature, one for each
> group, in K
> ref_t = 364.22 364.22 ; reference temperature, one for each
> group, in K
> ref_t = 366.04 366.04 ; reference temperature, one for each
> group, in K
> ref_t = 367.86 367.86 ; reference temperature, one for each
> group, in K
> ref_t = 369.69 369.69 ; reference temperature, one for each
> group, in K
> ref_t = 371.52 371.52 ; reference temperature, one for each
> group, in K
>
> P.S. I have realised that the longer the time spent in each replica the
> better is the sampling.
Maybe, depends who you ask. If the exchanges are actually independent from
each other and preserve the ensemble, then it doesn't matter. I think it is
clear that one can't usefully choose the period arbitrarily small, because
then the PE for successive attempts are correlated. But Sindhikara & co
disagree with me :-)
Mark
I should exchange replica every 1250 steps x
> exchange probability of 0.25 to get 10 ps of time spent in every replica.
>
> --
> Regards,
> Dr. Neha S. Gandhi,
> Curtin Research Fellow,
> School of Biomedical Sciences,
> Curtin University,
> Perth GPO U1987
> Australia
> LinkedIn
> Research Gate
> --
> 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