Subject: Re: [modeller_usage] Some theoretical questions
From: Modeller Caretaker <>
Date: Thu, 30 Jun 2005 10:45:55 -0700
Cc:
Dmitry Osolodkin wrote:
First of all, if I use optimisation of my structure (exactly as described in
Manual section 1.8.7.), do I need to minimise this system, for example, in
Sybyl to get better model? Or the model which I get is good enough for using
it without any further optimisation such as energy minimisation or molecular
dynamics?
You shouldn't need to do any further optimization - Modeller has already
done CG minimization and MD simulated annealing for you. In fact,
running an ordinary MD simulation of a Modeller model will often end up
making the model worse, as you're throwing away the homology-derived
restraints.
Second, Modeller can build number of models by one script. For what can it be
useful? And is there any difference between obtained models? What's the
reason for such a difference?
Each model is built from a different randomized starting model, so
they'll all differ slightly. You should build multiple models to allow
for the possibility that the optimizer gets stuck in a local minimum.
Third, is modelling of homomeric protein using the heteromeric template
authentical? All subunits of heteromer have the same folding as my homomer
and are homologous to my protein. Or can I build a model of one subunit of my
protein using the best template subunit and then build a pentamer using the
symmetry operations (rotation about C5 axis)? Is there such feature in
Modeller?
If you build one subunit, you'll be ignoring interactions between
subunits, so your model will likely be incorrect. You can build a model
of the whole system and use Modeller's symmetry restraints to ensure
that each subunit has similar structure. Modeller has no facility to
build new coordinates using symmetry operations.