[gmx-users] Help reg. EM

[sameer varma]

Hi,

I have been trying to energy minimize the structure of a protein complex 
but it has "exploded" in all the attempts that I have made so far.

THE FINAL GOAL: To create a homology model of a protein complex that has a 
known x-ray structure. In my case, the protein complex consists of 5 small 
proteins, each having close to a 100 residues. The x-ray structure clearly 
reveals the relative orientations of each segment.

WHAT WAS DONE: I independently created homology models of each of these 
segments. I placed these segments in their proper relative orientations. 
One can easily imagine that this would generate some steric violations. In 
order to get rid of these steric violations, I thought that I could use an 
EM protocol.

I tried using both steepest descent and CG algorithms, tried a range of 
step sizes (0.01-0.0001nm), tried several different force-fields (gromos96 
united atom, OPLS/gromacs all atom, gromos vacuum), tried using backbone 
freeze groups or backbone harmonic restraints (which is not advisable!), 
but in all cases, I find the same "strange" result. After just 1 step, all 
the segments are thrown apart. What I mean is, before EM, the segments were 
in close contact with each other. And after 1 step of EM, they are about 
15-20A away from each other.

What I fail to understand is, how could such a thing happen when the 
maximum displacement allowed by the algorithm in one step is bounded by the 
step size? Could this be a numerical issue?

Secondly and obviously: What can I do to fix this problem?

thanks,
Sameer.
URL: http://peptide.ncsa.uiuc.edu/~varma