[gmx-users] REMD, demux after 100 ns per replica

[Christopher Neale]

I have used demux.pl from gromacs 4.6.3 on files up to 1 us per replica simulated with gromacs 4.6.1. For my usage, the first column of both output .xvg files (replica_index.xvg and replica_temp.xvg) is fine up to 1 microsecond, then moves to scientific notation but, unfortunately, there are only five digits after the decimal so indeed some information is lost.

However, I was only attempting exchanges every 1 ps, so this is entirely consistent with what you are reporting.

Could it be that the issue is actually arising in the .log file? -- here are some snips from my mdrun log file:

Replica exchange at step 484096500 time 968193
Replica exchange at step 484097000 time 968194
Replica exchange at step 484097500 time 968195
...
Replica exchange at step 538320500 time 1.07664e+06
Replica exchange at step 538321000 time 1.07664e+06
Replica exchange at step 538321500 time 1.07664e+06

The good news is that, if this is true, this suggests that you can simply script a modification of (a copy of) your .log file and then run the standard demux.sh . This should probably be fixed in mdrun .log file at some point, although the necessary number of digits depends also on exchange frequency and duration of the run (i.e., gmx 4.6.1 seems to produce acceptable output for 1 us per replica with exchanges every 1 ps or, alternatively, should also be ok for 10 us per replica with exchanges every 10 ps, etc.)

In summary, I don't have a solution, but was able to confirm and slightly expand upon the bug report.

Chris.

________________________________________
From: gromacs.org_gmx-users-bounces at maillist.sys.kth.se <gromacs.org_gmx-users-bounces at maillist.sys.kth.se> on behalf of aperez <a.pervilla at gmail.com>
Sent: 16 April 2014 11:48
To: gmx-users at gromacs.org
Subject: [gmx-users] REMD, demux after 100 ns per replica

Hi all,

Today I have experienced a problem with the demux script applied in a
replica exchange simulation.
I have 16 replicas, and for each replica I have about 120 ns of MD.
When I executed the script demux.pl on the log file, I noticed that the
first column of the replica_*.xvg files had a problem after 100000 ps (100
ns).
I essentially got this:

...
99999.6       15   11   13    8    0   14    6    7    5   10    4    1    9
2    3   12
99999.8       11   15    8   13    0   14    6    7    5   10    1    4    9
2   12    3
100000        11   15    8   13    0   14    6    7    5   10    1    9    4
2   12    3
100000        11   15   13    8   14    0    6    7   10    5    1    9    4
2    3   12
100000        11   15   13    8   14    0    6   10    7    5    1    9    4
2    3   12
100000        15   11   13    8   14    0    6   10    7    5    1    9    4
2    3   12
100000        15   13   11    8   14    0    6   10    7    5    1    9    4
3    2   12
100001        13   15   11    8   14    0    6   10    7    5    9    1    4
3    2   12
100001        13   15   11    8   14    0    6    7   10    9    5    4    1
2    3   12
...

As you can see, after 100 ns the time step precision changes, loosing the
decimal place.
Does it mean that my 'demuxed' trajectories after 100 ns will experience
unexpected jumps? Because I saw this in my RMSD calculations...

This problem was fixed editing the demux.pl, in which this script reads
instead of the time ($log_line[6]), the frame multiplied by the dt -in my
case 2 fs- ((0.002)*$log_line[4]) from the log file, and also changing the
printf like this:

printf(NDX "%-20g",$t) ==> print (NDX "%-10.2f",$t)
printf(TEMP "%-20g",$t) ==> printf(TEMP "%-10.2f",$t)

However I am curious to know if you had already encountered such problem.
Thanks a lot for any possible comment,
Andrea





-----
Andrea Pérez-Villa
PhD student in Physics and Chemistry of Biological Systems
SISSA, Italy
--
View this message in context: http://gromacs.5086.x6.nabble.com/REMD-demux-after-100-ns-per-replica-tp5015885.html
Sent from the GROMACS Users Forum mailing list archive at Nabble.com.
--
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.