I've done homology modelling of a protein using modeller and itasser. Then to validate used 'rampage' and 'errat' validation tool. But, these two are recommending two different models are best. Now, how can I decide?
What you can do is comparative analysis of your model. Check out ramchandran plot and in that residues in the allowed region, i,e check ramchandran plot for the model obtained from modeller and ramchandran plot for the model obtained from itasser, also check that the models obtained from both have complete residues. Ramchandran plot will give you the stability of the protein.
Hello,
Since two servers are giving different results so basically then focus on the conserved residues involved in the model which you can easily find from the literature.
also you can check the active site residues if you know them these might help to identify how both models are differ.
one can also check for the secondary structure for arrangement of alpha helix and beta sheets and their total numbers.
i guess this might help.
best wishes,
Gaurao
When I make homology models I use numerous programs and compare the results. Recently I have found that PHYRE2.0 is really good modeling the proteins that I am interested in. Specifically, the PHYRE 1 server was terrible at modeling an adenylation enzyme, but when I submitted the same sequence a year later (PHYRE 2.0) the model was SPOT on. I do not care for MODELLER, but you will probably find that what works for your sequence does not work for others. HPRED is pretty good for all sequences that I have submitted. In all cases you must remember that your output is only a model that is to help you form hypotheses that need to be tested.
I does not hurt to throw a server generated model through a round of energy minimization to increase the quality of the model. As mentioned above, Ramachandran analysis is a pretty good idea... As is locating active site residues that you know are important for function. MOLPROBITY has quickly become the standard validation tool for protein structures (as implemented in PHENIX as well as the webserver: http://molprobity.biochem.duke.edu/index.php?MolProbSID=03ebfc282e23a556e8f375db42f254e0&eventID=15). MOLPROBITY also calculates a clashing score by adding hydrogens to the model and looks at potential residue flips.
Good luck.
Further the model you got from the servers should be validated in pdbsum or get the Q-mean score of them, also check whether the binding site or active site of the macromolecule is present in the modelled structure.
- Badges
- Science topic