[gmx-users] trjconv reference structure (bug?)

[Nguyen Hoang Phuong]

On Mon, 28 Oct 2002, Mikko Huhtala wrote:

> 
> I am trying to superpose trajectories to a reference protein crystal
> structure. The crystal structure is slightly different from the
> simulation starting structure, so I just converted the pdb to gro but
> did not solvate or add counterions etc. Thus the reference gro only
> has protein atoms and the trajectory has everything. When I give the
> name of the reference gro to trjconv with -s, trjconv starts, asks for
> lsq fit and output groups and then does nothing. The program never
> finishes and the size of the output trajectory file is 0. If I use an
> artificial reference gro file into which solvent etc. has been pasted
> after the protein crystal structure, trjconv -fit seems to
> work. Preparing the reference structure like this is not a big
> problem, just a pain in the fundament. (I am using precompiled Gromacs
> 3.1.4 from the rpm package on Linux.)
> 
> Mikko
> 

Hi, 

here is my understaning about the problem:

let's  go back to the lsf method:

suppose that you have two structures with the atomic coordinates XJ[i][j]
and  XJP[i][j] where i=1,2,3 (or x,y,z) and j = 1..N (number of
atoms) and you want to fit the structure XJ[i][j] to the reference
structure XJP[i][j].

To do that, one first builds the matrix U:

for(M2=1;M2<=3;M2++){
               for(M1=1;M1<=3;M1++){
                 for(J=1;J<=natom;J++){
                   U[M1][M2] = U[M1][M2] + W[J]*XJ[M1][J]*XPJ[M2][J];
                                   }
                                 }
                              }
where natom is the number of atoms in your structure and W[J] is the array
containing the mass of atoms. I think that in GROMACS, before doing the
lsf, one has to specify the natom of the group which you want to fit and
from that the memory of the array  XJ[i][j] is allocated. If that is the
case then one will have problem with the array XPJ[M2][J] because the
number of atoms in your reference structure are less than those in
XJ[i][j]. 

As you said the program never stops, this may be the diagonalizing routine
Jacobi which used to diagonalize the matrix U[M1][M2] is not converged
because of some strange things in this matrix.

Please correct me if I'm wrong. Thanks

Regards,

Phuong
-----
 


> 
> _______________________________________________
> gmx-users mailing list
> gmx-users at gromacs.org
> http://www.gromacs.org/mailman/listinfo/gmx-users
> Please don't post (un)subscribe requests to the list. Use the 
> www interface or send it to gmx-users-request at gromacs.org.
>