bias exchange and pbc: trivial or not

[Andrea Spitaleri]

Dear all,
i am getting crazy :) anyway I have been running a bias-exchange on 5
replicas (5 CV): 4 of them are distances and one is the antibetarmsd.
My system is composed by a peptide and a protein interacting each
other. So basically I am trying to see the undocking-docking
mechanism.
I am trying to see the trajectory and of course I am trying to remove
the pbc, which seems to be very hard (at least in this case ... never
mind).
from the xtc_with_pbc (directly from the simulation) I see that the
peptide is slowly leaving the binding site (using g_dist which takes
care of pbc).
Any options in the trjconv command (nojump, whole, cluster adn so on)
did not work to allow me to see the xtc trajectory in vmd or just
extract a frame. the peptide is jumping and it does not reflect at all
the analysis. I am wondering whether this problem is due to the
biasexchange simulation.

did you ever deal with this issue??

thanks for any suggestions

regards

andrea

--
----------------------------------------------------------------
http://sites.google.com/site/andreaspitaleri/
----------------------------------------------------------------
http://it.linkedin.com/in/andreaspitaleri
----------------------------------------------------------------

[Massimiliano Bonomi]

Hi Andrea,

I have not understood completely your problem. I suppose you are using gromacs.
When you say "trajectories from bias-exchange simulations",
you mean the continuos trajectories obtained by using the demux command of gromacs,
or the files as you get them from your simulation? In the latter case, these trajectories are not meant to be continuos,
due to the exchanges of configurations between replica.

Max

> --
> You received this message because you are subscribed to the Google Groups "PLUMED users" group.
> To post to this group, send email to plumed...@googlegroups.com.
> To unsubscribe from this group, send email to plumed-users...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/plumed-users?hl=en.
>

[Andrea Spitaleri]

Hi Max,
thanks for your reply. I will try to be clearer. Yes gromacs is the
engine and I am talking about trjectories from the simulation. Of
course they will not be continuous in fact I have also used the
demux.pl to demuxing them. So I have 5 demuxed trajectories too.

Now, the problem is that if I try to visualize the trajectory directly
from the simulation, beside the jumping which is expected, i see
broken molecules due to the pbc (as expected).
So on these xtc files (from the simulation which are not continuous),
I tried to remove the pbc in order to see the molecules in one cell.
Of course some jumping it is still expected (unless I guess I use the
demuxed trajectories) since I just removed the pbc but the xtc is
still discontinuous.
Then the analysis on the xtc_with_pbc (from the simulation) with
g_dist gives me a profile which does not correspond to the what I see
in vmd with the new xtc without pbc.
I mean from g_dist I see at time t1 a distance 10 whereas in the vmd
at time t1 the same distance is 1.8. that is saying to me that
something is wrong during my post-processing in removing the pbc,
isn't?
I just removed the pbc.
I hope to have been clearer :P

sorry for that

regards

andrea

2010/10/27 Massimiliano Bonomi <massimili...@gmail.com>:

[Massimiliano Bonomi]

The profiles from g_dist and the one from vmd differ only in few parts of the trajectory?

[Andrea Spitaleri]

I would not say few. some of them, randomly. I mean in some 'time '
differs a lot (1.8 vs 30 A).
do I need to worry or does it make sense to you?

thanks

and

2010/10/27 Massimiliano Bonomi <massimili...@gmail.com>:

[Massimiliano Bonomi]

Apart from units of measure (g_dist should be probably in nanometer and vmd in Angstrom),
It may be that with vmd you are not calculating distances with the minimal image convention.
When they differ, the value from vmd is usually very high?
What is the size of your cell?

Max

[Andrea Spitaleri]

Yes I am considering the difference in units :P

> It may be that with vmd you are not calculating distances with the minimal image convention.

in fact this is case I guess. but removing the pbc I should not have
this problem but probably trjconv is not doing this correctly (which
is not new to me). Now I have also the bias so I thought that there is
something more complex.

let's say i have in g_dist 0.2 nm and in vmd it say 30 A (if i
remember well the number but the order is this) my cell is 6 nm x 9 nm
x 6 nm

[Massimiliano Bonomi]

With trjconv you just make your protein whole again.
Then you can calculate a distance between two atoms with or without the minimum image convention.
I guess vmd does not use this convention.

Max

[Andrea Spitaleri]

Ok. in fact i trust gromacs not vmd. I will find a workaround to see
some structures.

thanks for now

[Xevi Biarnés]

Hi Andrea,

now I understand better your problem. you have two different molecules (protein + peptide ligand) jumping across boxes in the xtc files. right?

so, as you say, vmd does not take into account pbc when computing distances. so the big distances you get (>30 A) should be corrected by the box size. and your actual problem is not in getting the right distance (which you can get from g_dist), but visualizing the structures toguether in VMD.

in VMD, you can put back together the ligand and the protein in the same box, by applying this script I wrote a time ago. Once you have finished, you'd better save the trajectory (from VMD) into a new file to keep the changes.

hope it helps.

Xevi.

[Xevi Biarnés]

sorry... the attached file

let me know if it works to you.

Xevi

[Andrea Spitaleri]

Hi Xevi,
thanks a lot. Yep that is the question. Sorry for not being clear previously.
This "swiss knife" is very handy!! It does what is wrote in :)

Basically I tried in different ways and these are the best solution:

##method A xtc FROM BIASEXCHANGE
1. trjconv -s -f -pbc mol -center -n index ## select either first the
protein and then output the complex or complex for both selections;
2. open the new xtc in vmd and use arrange_objects 0 all all (other
selection not seems to be right)

##method B xtc AFTER DEMUX
1. trjconv -s -f -pbc mol -center -n index ## select either first the
protein and then output the complex or complex for both selections;
2. open the new xtc in vmd and use arrange_objects 0 all all (other
selection not seems to be right)

both are working getting of course two different results. The
distances now are visualized correctly.
However, there are still some frames (very few) when the peptide is
split in 3 (attach). basically the peptide is in the corner and so it
is divided in 3. I am checking with the demux xtc what's going on.

For now thanks. I hope this will be useful for others

and

2010/10/27 Xevi Biarnés <xevib...@gmail.com>:

[Xevi Biarnés]

great!!!

usually,

arrange_objects  "top"   "protein"    "not protein" 

should do the thing. anyway, if it works, don't care!

regarding the broken molecule, you could give a try to    trjconv -pbc whole