I just want to add a comment to what Mickael just told you.
Basically, Modeller use spatial restraint derived from the homology, which means that it will derive the restraint for the side chain from the residues in the template with which is aligned.
Modeller has a table for equivalent position between residues: example Alanine and Lysine will share as equivalent backbone atoms and the Calpha . This means that if your template is Ala and your target is Lysine, only homology restraint concerning backbone + Calpha will be derived from the Ala and applied to the Lysine.
That implies that your Lysine side chain will be optimize with non homology derived restraints (i.e. those from template),and so more dependent on the starting conformation, the environment and, the potential your are using.
At the opposite if you check the identical residues you should see less fluctuation.
Eric.
Liu,
Your problem deals essentially with side chains. You have to know that all
programs of homology modeling have difficulties for prediction of side
chains and results are often bad...
Basically, at the biginning, programs choose conformation of side chains
of templates in conditon that the same residuse are aligned. If it is not
the case, the side chain is placed and is minimized. So, it is logical
whether your side chains are different between your models because the
minimization depends on the local environnement.
I hope you have understood what i had tried to explain...
With regards,
Mickaël
On Wed, 24 Mar 2004, Binbin Liu wrote:
> Hi there,
> I got one hunderd models. The side-chain torsion angles of the interesting
> residues are vary. Does anybody know why. Does MODELLER calculate the torsion
> based on some rules or it is by chance.
>
>
> Cheers
>
>
> BinBin
> _______________________________________________
> modeller_usage mailing list
> modeller_usage@salilab.org
>
http://salilab.org/mailman/listinfo/modeller_usage>
_______________________________________________
modeller_usage mailing list
modeller_usage@salilab.org
http://salilab.org/mailman/listinfo/modeller_usage