Dear Susy,
You describe two things: making a homology model and designing mutants.
As Oliv pointed out, the first step is getting the alignment right. Even with a multiple sequence alignment and knowledge of the template that can be quite difficult. You may need to make several models for alternative alignments.
When validating your model, packing is IMO the most useful thing to look at. Rosetta-holes is very good at finding poor models. What_check also has very good packing analysis and also flags unsatisfied hydrogen bond donors/acceptors which may help a lot. There are also some compound scores such as the MolProbity score and Q-mean which may help you to find the best model from a set.
The quality of the template is also important. Get the best template or use more than one. If the sequence identity isn't too low you should also use templates from pdb_redo. They helped (a lot?) in the previous CASP.
If you want to analyse single mutants you can try to work from simple principles such as minimizing the loss of torsional freedom of the main and side chain (entropy) or optimizing the gain of water entropy when the protein folds. Also minimising the loss of hydrogen bonds upon folding is quite useful. Just don't expect miracles.
Cheers,
Robbie Joosten
Date: Tue, 8 Feb 2011 10:52:09 -0800
From: [log in to unmask]
Subject: Re: [ccp4bb] quality homology model
To: [log in to unmask]
Hi Susy,
Before going into details: you have to get the alignment right.
After that, you may look into packing quality and geometry of your
models.
Regarding the parameters you are interested in: I talked to a guy
from the Sali lab (Modeller) about quality assessment of homology
models. He recommends to compare the gromos or anolea profiles per
residue (or whatever packing score your software uses) to the
template. In case there is gap: pay attention to the alignment!
Regarding cut-off values (what's ok and what not) you may want to
look into published work on your modeling software and/or contact
the authors directly.
Another way to look at model quality, and maybe more significant, is
to check geometry (clashes, rotamers, bond angles etc.). You could
use MolProbity for that. The supplementary material of the following
article contains a good example (Table S1) on the evaluation of a
trimer interface:
http://www.ncbi.nlm.nih.gov/pubmed/20457942
Good luck!
Oliv
On 02/07/11 01:13, Susy Rao wrote:
Dear Community,
I have a question regarding protein model quality for
introduction of point mutations which should increase
solibity/stability of my protein of interest.
I plan to do homology models so that I can check the most
promising amino acids.
Which quality/resolution of the model do I need?
Which parameters would you check (gromos, anolea), and what
would you use as cut-off values?
At the moment I thought, that it could be interesting to
model a homologue of my protein and check the RMSDs with the
crystal structure.
Maybe you know some good literature, describing this?
Thank you
Susy
--
Oliv Eidam, Ph.D.
Postdoctoral fellow
University of California, San Francisco
Dept. of Pharmaceutical Chemistry
1700 4th Street, Byers Hall North, Room 501
San Francisco, CA 94158 - Box 2550
Phone: 415-514-4253
Fax : 415-514-4260
Email: [log in to unmask]
