On Thu, Sep 23, 2004 at 05:34:35PM +0300, vivek sharma wrote:
> I went through the model.top script and and tried to understand that how it
> works. But there are few problems in understanding that script, please guide
> me for these mentioned problems.
I suggest you read the original Modeller paper and its references, where
this and other points are discussed in more detail:
http://salilab.org/pdf/029_SaliJMB1993.pdf
> 1. In OPTIMIZE process (VTFM optimization), the addition of restraints is
> governed by PICK_RESTRAINTS which in turn depend on SCHEDULE file (3rd
> column, NRNG), meaning thereby that only those restraints will be taken into
> consideration which applies on atoms, which comes in category as specified
> by Schedule file (3rd column). Although in .log file I found restraints are
> added step by step (N_SCHEDULE step) during optimization , are the atoms
> also added continously, or just explain a bit how this profile goes?
All atoms in the model are optimized - it's only the restraints which
are added gradually.
> 2. Out of curiosity I ask this question, if all restraints and all atoms are
> taken into consideration at a time, and then optimization is done, will that
> be better/efficient than the method it takes in model.top?
The fewer restraints, the simpler the energy surface, and the easier it
is to find a global minimum. Modeller's optimizer will probably have
trouble building a good model if you start with all restraints enabled.
Of course, you can write a simple schedule to test that, if you like.
> 3. Also, I observed that this addition of restraints (for every N_SCHEDULE
> step) during optimization increases the value of objective function , and it
> finally stops at a higher value, will that value be considered final one or
> various such consequent optimization cycles should be carried out to finally
> lower down the value?
The objective function is a function of the defined restraints, and so
it will increase every time additional restraints are added. You can't
meaningfully compare values of the objective function between different
sets of restraints, or different models.
Ben Webb, Modeller Caretaker
--
http://www.salilab.org/modeller/
Modeller mailing list: http://salilab.org/mailman/listinfo/modeller_usage