On 03/28/2012 03:32 AM, Sumedha Roy wrote:
I tried modelling a new complex based on a template complex with the same DNA ligand but a few mutations in the protein. Initially I did the modelling the usual way and got good detectable RMSD (about 0.6) with the new mutated PDB (just the new protein, didn't have the DNA) and the original complex. However when I tried to do the modelling considering DNA as a ligand (changed all DNA ATOMs to HETATM and modified alignment file with 9 appended '.'s - as my DNA is 9bp long), the new mutated complex, is nearly SAME as the template complex with minor changes in coordinates. (0.1 RMSD) I don't understand that if I am using the same template and mutations, why is there so much of a difference (nearly 0.5 RMSD) between the two new complexes (with and without the DNA).
This doesn't sound that surprising, given that you are modeling two different things (in one case just the protein; in the other, the protein plus DNA). As Ashish has already pointed out, likely you are also measuring different RMSDs in the two cases. When you add ligands to your model, Modeller adds restraints between protein and ligand to maintain the conformation. Those restraints will pull on the protein atoms, of course, which is an effect you don't see when you don't have the ligand.
RMSD between model and template is a very poor measure of model quality (unless your sequence identity is very low or your alignment is terrible, the RMSD will always be quite low, since Modeller builds models that resemble the template by construction). A good assessment method like DOPE is far superior.
The nonstd_restraints() method is called automatically during model building, so you don't need to call it yourself. It adds the restraints I described above.
Ben Webb, Modeller Caretaker
--
http://www.salilab.org/modeller/
Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage