[gmx-users] Re: Saving trajectories of a particular group within the system.

[Mark Abraham]

Junqiao Lee wrote:
> Hi Jeroen, Tsjerk,
> 
> Thanks for the help. I've tried both ways and both seem to work; with
> the dash or the underscore...
> 
> Funny thing is that the grompp output seem to indicate that there are
> 2 XTC elements (Protein rest) when I've set xtc_grps = Protein.
> 1 XTC element (rest) when not setting xtc_grps to any specific groups
> (i.e. default).
> 1 XTC element (Protein) when setting xtc_grps = Protein SOL or Protein
> Non-Protein ...etc.
> Again, I'm currently using Gromacs/3.3.3

Your use of "seems" suggests to me that you haven't actually done an 
experiment controlling all variables except the spelling of the name of 
this option and the names of the groups you're giving. As such, your 
observations don't help anybody much :-)

You can use gmxcheck to compare pairs of run input files, which can be 
useful here.

> However, looking at the size of the output-files, xtc_grps = Protein
> seem to be correct.
> The other two generated much bigger .xtc files.
> 
> As previously indicated, I'm only interested in the various
> structural conformations adopted by the protein-molecule as it samples
> across the potential landscape during a very long simulation (> 10ns
> perhaps in steps of 1ns). I intend to output the coordinate
> trajectories at very short time-intervals of 50 fs. i.e. Infact, I
> only need the coordinate data of the protein during the
> virtual-cooking period, and not the velocities of the protein, and
> none of the data of the solvent (+ counter ions) molecules.
> 
> The relevant section of the current .mdp input-file settings I've made
> is as follows:-
> ===================================
> ...
> ...
> 
> nstcomm                  = 10
> comm-mode                = Linear
> 
> xtc-grps                 = Protein
> nstxout                  = 10000
> nstvout                  = 10000
> nstfout                  = 0
> nstxtcout                = 50
> 
> nstlog                   = 1000
> nstenergy                = 1000
> 
> nstlist                  = 10
> ns_type                  = grid
> 
> ...
> ...
> =======================================
> 
> I'm wondering if by setting xtc_grps = Protein, am I to expect the
> generated .g96 and .xtc file (after mdrun have completed the full
> simulation) to contain data particular to the Protein only, which was
> what I was hoping for? However, loading the .g96 file into VMD to

The file produced by mdrun -c contains the whole simulation system 
regardless. The file produced by mdrun -x contains the groups, data 
types and frequencies indicated in the corresponding .mdp file options.

> visualise it, the solvent molecules (all ~7800 of them) was found to
> be still there. And loading the .xtc file to the molecule, I can see
> both the protein and solvents were moving, which seem to indicate to
> me that the .xtc file actually contains data for the whole system
> basically (instead of just the protein). This was what had confused
> me.
> 
> However, I've just realised that the .xtc file does seem to be much
> smaller, after performing the short test runs to compare.
> 
> Just to verify, by setting xtc_grps = Protein, what should I expect
> with the various files (.xtc, .trr, etc.). Should I expect the
> simulation to throw out trajectories of the Protein group only to the
> .xtc file at time-interval set by "nstxtcout" (fs). While "nstvout"
> and "nstxout" sets the data to output to the .trr file, would that be
> correct? Or is there some misunderstanding on my part. I've read the
> manual as carefully as I can, but as a human, I might still make
> mistake or interpret what it say wrongly. Hencefore, I need to
> cross-check before I initiate the (much longer) real simulation; since
> what I've seen with VMD looked a bit "dodgy" as mentioned above.

The lesson here is to read carefully, and experiment carefully. If you 
don't know what you've varied, then your observations will be confused, 
just like in real laboratory science :-) In the experience of a lot of 
competent people, all these options work as described, and you need to 
be on very strong ground to suggest they don't! :-)

> I've set "nstxout" and "nstvout" to a much larger time-interval, as
> the .trr file-size tends to blow up rather quickly (because its not
> compressed?), 

Yes.

> and I'm assuming those two parameters are with regard to
> the .trr output and not the .xtc output during the simulation.

Sure, like the documentation says.

> Does my .mdp file above, correspond to what I'm intending to achieve
> (as described above - i.e. saving only the trajectories of the protein
> during the simulation)? Any suggestions for perhaps better way of
> doing it would be helpful too. =)

You're doing it roughly in the manner intended - unless you've supplied 
an index file with the Protein group erroneously including the entire 
system. At least for normal systems built with pdb2gmx, the 
implicitly-generated Protein group will work as you want it to here.

> Hope I've provided sufficient information for you to "go with" and
> that its not too confusing to read and get what I'm trying to say,
> English isn't really my mother tongue and sometimes I can be a bit
> confusing no matter how hard I try to articulate myself
> properly/clearly. Further, hope I've not annoyed anyone too much by
> asking stupid questions. I'm new to the MD thing and I've just started
> working on this as a short 10-weeks internship, after having completed
> my degree in Nanotechnology (i.e. Physics, Chemistry major), as I find
> this project interesting.

Well, good luck with that.

Mark