[gmx-users] g_dist as an alternative for g _saltbr

Justin Lemkul jalemkul at vt.edu
Fri Oct 17 21:56:43 CEST 2014

On 10/17/14 2:13 PM, Diogo Martins de Sá wrote:
> Dear Justin,
> thanks for your reply. To start with, here are the links where you
> mentioned "looping calculations"
> (http://article.gmane.org/gmane.science.biology.gromacs.user/59569/match=g_saltbr+looping+calculations)
> and "track persistence of interactions"
> (http://www.researchgate.net/post/Can_anyone_help_with_salt_bridges_with_gromacs)
> Since g_saltbr is dumb in its main function (and should actually be
> called something like g_electrostaticInteractions), I've decided to
> tackle this question in the following manner:
> 1. I must use gromacs in way where I find all the SBs in my system, and
> just those.
> 2. I must use gromacs to observe the distance of these SBs as a function
> of time.
> This is how I intend to do it.
> 1. Acording to Kumar & Nussinov 2002 (Close-Range Electrostatic
> Interactions in Proteins), salt bridges must have at least a pair of
> sidechain functional-group nitrogen and oxygen atoms within 4 ä
> distance. So I will use g_select to create groups of ASP's & GLU's
> sidechain Oxigen and respective hydrogens and HIS's, Lys's & Arg's
> sidechain Nitrogen atoms and respective hydrogens. Why include the
> hydrogens? Because I'll need them to use g_hbond. Why use g_hbond?
> Because of the matrix. Why is the matrix so important? Because of your
> (Justin's) script that shows the percentage of time that a given
> hydrogen bond existed during the trajectory:
> http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin/scripts.html.
> So I will run g_hbond with -r 0.49 (0.35 from hydrogen bond + 0.14 which
> is the furthest distance a covalently bonded hydrogen can be from
> nytrogen, this distance is 0.12 for oxigen. Hydrogen bonds pull the
> covalently bonded H away from atoms and this is directly proportional to
> the strengh of the bond). After this process, I will have all the salt
> bridges and the percentage of their appearence through the trajectory
> (thanks to your script). So I now I know which ones I should take a
> closer look.
> 2. To measure distance as function of time, I can use either g_bond
> (with -d) or g_dist. If I get few SBs after step 1, I will use g_dist,
> because then I will make individual groups. If I get more than a few
> SBs, g_bond can organize each one and I have to make only two groups.
> What do you (and any other user reading) think about this strategy?
> I do have one problem, I never used g_select and I'm having some
> difficuly understanding the syntax. Does any one know how I should call
> g_select so I create index groups like I mentioned in step 1??

Probably something that is even easier is to use trjconv to strip the groups of 
interest using a single index group (i.e. functional groups of interest), then 
use tpbconv (gmx convert-tpr in 5.0) to make a matching .tpr, then just use 
g_saltbr -t as needed.  Saves a lot of work, and you're only ever looking at 
possible groups of interest.  That circumvents the problem that g_saltbr doesn't 
accept an index group.  trjconv and tpbconv do :)



Justin A. Lemkul, Ph.D.
Ruth L. Kirschstein NRSA Postdoctoral Fellow

Department of Pharmaceutical Sciences
School of Pharmacy
Health Sciences Facility II, Room 629
University of Maryland, Baltimore
20 Penn St.
Baltimore, MD 21201

jalemkul at outerbanks.umaryland.edu | (410) 706-7441


More information about the gromacs.org_gmx-users mailing list