[gmx-users] pulling code

[chris.neale at utoronto.ca]

Dear Adam:

I have a number of questions listed below. But first, I think it would  
be useful if you explain exactly what you want to accomplish and how  
you tried to do this with non-modified code and what problem you ran  
into that caused you to modify the code. After you outline that,  
please address the following questions:

1. Why not use one terminus as the reference group and the other  
terminus as the pull_group0 ?

2. How did you modify the code to use the shortest vector? I was under  
the impression that the code did already do this (shortest vector  
accounting for PBC) and that this was actually something you wanted to  
avoid.

3. Why can you not just use A-B even if A is positive and B is  
negative? If you are trying to avoid determining a PBC-wraparound  
distance, then it seems like A-B does give you exactly what you want  
as long as the entire protein is inside the unit cell.

4. If you really need to know the box size, t is available in the code  
at runtime. Why can you not just use the appropriate variable?

5. How are you fixing the non-terminal groups that you say need to be  
fixed? I don't understand why you can't fix groups with position  
restraints and the pull with the pull code. Also, why do you need to  
fix certain non-terminal groups?

6. This confused me a lot: "I don't think I want to freeze or restrain  
the groups that need to be fixed, because they must still be allowed  
flexibility to twist and deform--I only want to roughly fix their  
centers of mass." ... so then how did you fix them?

Chris.

-- original message --

Hi all,
I am trying to measure the energy curve of unwinding the SNARE protein
complex.  This requires pulling the C-termini of two helices apart from each
other, while fixing certain other parts of the complex.  To do this I
excluded the reference group, so that all groups were pulled to absolute
coordinates.  However, as the protein was centered in the box, and because
the box center is where the coordinates transition from -boxlength/2 to
+boxlength/2 when u7. Why do you need to fix certain non-sing pbc (ie.  
the dateline), every time a group passes
through the center of the box in any of the 3 dimensions, it is suddenly
yanked violently through the box because it thinks it is now a full
boxlength away from its destination.  I modified the pull code to take the
shortest vector between the current position and destination, which seems to
fix this problem.  I also modified g_wham so that it can take the difference
between two groups' positions as the energy curve coordinate (since several
groups need to be fixed, neither of these two groups can be used as the
reference group).  However, as I use pressure coupling, the box dimensions
change during the simulation, and if one group has a positive and the other
a negative coordinate because they straddle the center of the box, the box
dimensions must be known at each timestep to get their actual separation.
  Is there an easier way to handle this type of simulation?  I don't think I
want to freeze or restrain the groups that need to be fixed, because they
must still be allowed flexibility to twist and deform--I only want to
roughly fix their centers of mass.  Any suggestions would be appreciated,

Thanks,

Adam