submit it to FG-MD http://zhanglab.ccmb.med.umich.edu/FG-MD/

Our just-published protein structure refinement method can take a structure and improve its accuracy relative to the native consistently, combining concepts from machine learning, optimization, and simulations.

Feel free to submit it to the Princeton_TIGRESS method for refinement:

http://atlas.princeton.edu/refinement/

Here is the associated paper:

http://onlinelibrary.wiley.com/doi/10.1002/prot.24459/abstract;jsessionid=BC6E3B978A3F881711BEB86246A963EA.f01t03

You could try to do a simple minimization of your structure e.g. with Amber. Why are there clashes in the first place?

thanks for your responses, but still my problem is not solved.

Either the above programs do MD or use SVM to refine protein structure but I want to only change the values of phi and psi torsion angles of only those residues which fall in disallowed region of ramachandran plot so as to remove steric clash (and not do any MD or any other modification in protein structure).

@Alexander Rose --> Amber is not able to minimize above attached structure because of such steric clash thats why i asked for above help. For a protein/peptide sequence I got some phi and psi angle values for each residue with some initial guess (say if predicted secondary structure of that residue is helix then assign phi/psi of -57/-47). With these phi and psi values I made initial structure pdb file using amber. but because there can be steric clash as shown in the above example, i want to minimize that structure. Amber failed to minimize above structure due to serious steric clash in the backbone. This is the reason i want to make all phi psi angles to lie within ramachandran allowed region so as to avoid such serious steric clash.

Your clash seems to be quite long range (the clashing atoms must be at least six residues apart), so simply regularizing the Ramachandran values, which are defined by local steric constraints only, may not solve the problem.

@Matthew Conroy --> I understood your point which I didn't thought before. It seems even if phi and psi values occur within ramachandran allowed region, still this clash can occur. But then what can be the alternative to solve my problem? I just want phi psi values which doesn't produce any clash.

On the other hand, moving one torsion angle may (and probably will) swing the whole of that loop out of the way. I think Swiss-PdbViewer (google spdbv) enables you to dial in backbone torsions and change a model.

The bigger question though, is what do you hope to learn from this model?

@Matthew Conroy --> I want something that can be automated using any command line script. Swiss-PDBviewer or any other GUI based program doesn't solve my problem.

If you only care about the backbone you could remove all sidechains, i.e. mutate all residues to glycine, and then try to minimize this backbone-only-model.

You could try the minimize input file for cns (http://cns-online.org/v1.3/). It is extremly fast and I had some good results with morphed structures. It's also quite easy to automate,using scripts (Linux only, though, AFAIK. You can contact me, if you need an example for a script).