If your model was constructed on a high identity score, and if the backbone atoms seem to be stable you can use it.

I would suggest you to go for FEL analysis to sample the low energy conformations..in case if u have done md with gromacs, you can use g_sham tool http://www.bevanlab.biochem.vt.edu/Pages/Personal/justin/Scripts/sham.txt

If RMSD is stable at 4 nm (GROMACS default unit), then you cannot say how much real difference in RMSD between conformations during 100 ns to 200 ns. 4 nm is a big number for RMSD!! It also means that structure is significantly changed from the starting one and probably modeled structure is wrong.

As suggested above, you can do PCA to monitor free energy landscape. However, if RMSD is 4 nm, PCA might not be suitable because most likely fitting will not work with such a high RMSD.

Thank you Dr.Umashankar and Rajendra for you inputs.

@Rajendra There are few comments:

1. I missed out the RMSD units-- the units are in the Angstroms not in nm. The modeled structure is showing stability from 70 to 100 ns to be precise.

2.The modeled structure was checked for authenticity prior to MD , whch showed 91 % residues in the allowed region of Ramachandran plot .

My query comes back to same can i extract the coordinates at 100 ns or 200ns for furthur study?

Thanks in advance !!

Still, 4 A is a large RMSD for small to intermediate system. I do not know the system size, therefore, I cannot say much about RMSD.

I would suggest to construct a 2D free energy landscape from PCA and try to extract a conformation with lowest free energy from largest cluster.

Thank you Rajendra !!