Re: MULTIPLE/SINGLE templates for modelling.

[Procter James <>]

I can offer you some advice, Mr Sridhar, but I cannot call myself
an expert :

> 1. In homology based modelling is it good to use single template or
> multiple templates?
In general, yes, because the templates provide information on the parts
of the structure least likely to vary with changes in sequence. It is 
not such a good idea to use templates with huge variations, as is typical
amongst structures from remote homologs. It can also be unhelpful to
use every homolog - this adds noise to the structural data.
The most important factor is that the structures are all of good quality,
and are as close as possible to the sequence to be modelled, particularly
in the functional and/or highly conserved parts of your sequence.  

> I have a cytochromep450 protein which shares maximum homology with another
> protein of same class . But there were some missing residues in the template.
> When used for modelling the region corresponding to this missing region had
> bad environments in Verigy3D. I went on for loop modelling for this region.
> But is this a good approach ?
Loop modelling is the logical approach if no reliable template can be
found for these regions. By default, modeller does not optimize the geometry
for unaligned residues of a sequence - so these parts do not score well
in a structural validation program. Loop modelling will improve their
geometry, but is not guaranteed to give you the perfect answer!
 
> I feel the alignment itself should take care of everything. I heard about
> using multiple templates in such cases. How is this approach followed and
> how good is it ?
The alignment is of paramount importance (as you would read if you look
at the archives of this mailing list). For multiple templates, one must
take care that multiple sequence alignment artefacts are not introduced.
Whilst a good MSA will provide the best alignment data for a homology
model, sometimes there are locally inconsistent residue mappings, sometimes
within conserved regions of secondary structure, because one or more
of the proteins has a region of low homology. If these are not filtered out,
ultimately by a little trial and error, then they can severely reduce the
model quality.
> --------------------------------------------------------------------------
> 
> 2. I need to model some point mutations into this same protein sequence.
> What is the best way of introducing mutations. Should the mutations be
> introduced in the sequence and then modelled seperately, or should they be
> introduced in the model of the wildtype ?
You should try them both. If you introduce a single mutation into the wt. 
that catastrophically affects its structure, you are more likely to see
this from the model. Changing the sequence that you are modelling only adds
a little more noise into the final structure - which may not be correct to
start with.

> I want to do molecular dynamic simulations on Wildtype and Mutants to
> compare and contrast the wildtype and mutant structures. What effect would
> both these approaches have on the correctness of my results.
Again - this is another question, that can't be easily generalised.
I think the answer above still applies - and you should observe some
meaningful behaviour far earlier in the wt. mutant's simulation than
in a model that is already full of noise.

This was supposed to be a quick reply - but I hope it helps.
j.