In NAMD one can use the tclforce call to define a good set of collective (CV) coordinate. I seems that you know which conformational changes are needed to onvercome the free energy barrier which separated both states. This is a plus, what you need is simple the correct script to perform a target MD, well a bit of TCL scripting using the abovemention calling in NAMD can do much for you. Another option, which is simple and do a lot of thing is PLUMED, which has support for namd and gromacs as well. Try to install and look for the right section about CV and its manipulation. best

Dear Nivedita, 

Unfortunately, TMD is not implemented in Gromacs. You might want to give a try to NAMD!

TMD is available in CHARMM, I have attached a link to its documentation. I have tried it on a protein same size as your using EEF1 implicit solvent model also available in CHARMM.  

http://www.charmm.org/documentation/c34b1/tmd.html

Yes, you can do it. The portion of the protein you want to subject to MD consider that as completely movable, and surround 3-4 A area as tethering and subject both parts to MD, then protein will not be denatured. However, it may not be easy in Gromacs, but use regular MD program from Accelrys, Amber etc.

In NAMD one can use the tclforce call to define a good set of collective (CV) coordinate. I seems that you know which conformational changes are needed to onvercome the free energy barrier which separated both states. This is a plus, what you need is simple the correct script to perform a target MD, well a bit of TCL scripting using the abovemention calling in NAMD can do much for you. Another option, which is simple and do a lot of thing is PLUMED, which has support for namd and gromacs as well. Try to install and look for the right section about CV and its manipulation. best

As Adolfo pointed out, You can compile GROMACS along with PLUMED and use the path collective variables (there is one for RMSD that is equivalent in its spirit to what NAMD has, and there are other ones that may be more useful for your system in particular) to do steered (targeted) molecular dynamics.

http://www.plumed-code.org/

Good luck!

It is indeed easily possible to do this in Gromacs with essential dynamics. Put open and closed configuration into one pdb file and make a PCA with g_covar. This will give you exactly one eigenvector that connects open and closed state. then call make_edi to generate an input file for essential dynamics. Choose the "radcon" option with your target configuration as reference. Then start mdrun as usual but with an additional -ei argument. See make_edi -h and  mdrun -h for details. Hope that helps. Ulf

As Ulf hensen said we can get one eigenvector that will connect open and closed states. Are you trying to say that keep open and close pdb file into one file and simulate the protein and go for PCA analysis which will give you exactly one eigenvector which will connect open and closed states. But is it really possible bcoz  the protein molecule itself having more than 2000 residues?

You don't need to carry out MD to get the connection vector between open and closed state. Just concatenate open and closed state (cat OPEN.pdb CLOSED.pdb >merge.pdb ... take care that you have only one header in that merged file), then do a PCA (g_covar -s OPEN.pdb -f merge.pdb) and you will obtain one eigenvector in eigenvec.trr that is your connecting vector with an eigenvalue corresponding to the MSD between open and closed state. Restrict yourself to the C-alpha atoms if your system is very large. In your case, your eigenvector will be 3*N=6000 dimensional. This ought not to be a problem on a modern workstation.