Re: [modeller_usage] manual restaints in muultiple template modeling?
Subject: Re: [modeller_usage] manual restaints in muultiple template modeling?
From: Modeller Caretaker <>
Date: Tue, 02 Dec 2008 09:43:59 -0800
> I want to build a model using two templates. One of the templates
> have >70% sequence homology while the other has <35% sequence homology.
> To the extent of my knowledge, automodel gives more weight to the
> template with more sequence homology. Is there any way I can manually
> assign more weight to the template with less sequence homology, at some
> of the regions? If there is, where can I get the details, as to how to
> do it?
There is good news and bad news:
The good news is that Modeller does not weight the templates simply by
their overall sequence identity. Instead, they are weighted by their
*local* sequence similarity, over a sliding window of 5 residues. So for
example, even if a template has a lower sequence identity to your
target, Modeller will still "pick" it in regions where it is locally
more similar to the target than the other template. The weighting is
also not a simple "take 40% of template 1 and 60% of template 2" -
Modeller uses its internal database of triple alignments to build
reasonable restraints when two templates suggest different feature
values. See the 1993 Modeller publication for more details.
The bad news is that since the weighting algorithm is not a simple one,
you can't adjust it, at least not without a good understanding of the
See also http://salilab.org/modeller/FAQ.html#2 for the extreme case
where one template should not be used at all.
Ben Webb, Modeller Caretaker
Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage