using dgsa with missing atoms

Pochapsky Rotator tommyrot at ERNST.NMR.BRANDEIS.EDU
Wed Apr 9 11:51:06 EST 1997


OK -- let me start by saying that I am an inexperienced xplor user.
Having said that, my problem is the following:

	I am running xplor with dgsa.inp on a series of files that were
previously generated using dg.inp.  These pdb files contain only
hydrogens.  

	The problem:  in the section where unknown atoms are taken care of by
reading in a template file and least squares fitting is performed
(quoted below), the program terminates execution during the LSQ fit
about half the time. The error message is: rotation matrix is not
unitary.

	My suspicion: the template file that I'm reading in and comparing to
contains the same hydrogens as the pdb file, but the positions are not
always (in fact, not usually) the same.  My best guess as to what might
be happening here is that xplor is reading in the template file, which
is *supposed* to have all the known atoms at the same positions, and
performing a least squares fit to orient properly when it supplies the
coordinates of the missing atoms.  I think that if this is what's
happening, the likely reason for the crash is that my hydrogen
coordinates are just too far from the ones in the template file.  (the
template file that I'm using is a product of a previous structure
determination on the same protein, so it should be reasonably
approximate but by no means correct). 

	What I need to know:
	a) am I right about what's happening
	b) if I'm not, does anybody have any better ideas?
	c) if I am, I'm not sure that this is the best way to do this.  It
would seem to me that the ideal would be to use the hydrogen positions
that I computed and then supply the missing atom positions based on
amino acid construction.  Minimizing from this point would seem to make
more sense, as it avoids biasing the calculations by using structures
from a previous determination with different constraints.  The LSQ step
could then be omitted and the sa run in a fashion similar to an ab
initio sa.  Does anybody know a good way to make such a file (keeping in
mind that I have a series of 10 hydrogen pdbs that I need to convert)?     

	Due to peculiarities of the accounts here, etc., please mail replies to
jleah at silver.lcs.mit.edu as well as posting them.  Thanks in advance for
your help!


Jennifer Heymont
jleah at silver.lcs.mit.edu





{* ================== Create local ideal geometry by template fitting;*}
     {*                    this also takes care of unknown atoms.
*}
           
      set message=off echo=off end
      coor copy end          {*Store current coordinates in comparison
set.*}
      
{====>}                 {*The user has to supply a template coordinate
set.*} 
      coor @@generate_template.pdb  
                  
      for $id in id ( tag ) loop fit              {*Loop over residue
tags.*}  
                  
         coordinates                 {*LSQ fitting using known
coordinates.*}
            fit select = ( byresidue (id $id) and not store1 )
         end
                      {*Store fitted template coordinates for this
residue.*}
        coor copy selection=( byresidue (id $id) ) end
      
      end loop fit
            
      coor swap end
      set message=on echo=on end



More information about the X-plor mailing list