Objective Function vs. Molecular PDF
Hello,
In the end of the log-file from a MODELLER (8v1) run there is a summation of the quality of the model:
>> Summary of successfully produced models: Filename molpdf ---------------------------------------- A.B99990001.pdb 2355.08740 N.B99990002.pdb 2376.38696 ... ...
My question now is if these values really are the molecular pdfs or if it is the objective function?
The reason I am asking is, that we have built homology models of the same protein two times using the same sequence alignment and template pdb-files only changing the refinement level - one is the default the other is md_level = refine.slow_large.
In the second run the values in the log-file range around 9000. If the value really represents the mol PDF this is good, but if it is the objective function, the first models (default refinement) with the lower values are much better...
Another question: What is the default level of refinement actually? I have not been able to deduce this from the manual.
I hope someone can help me with this:-)
Thanks in advance Leyla
Leyla Celik wrote: > In the end of the log-file from a MODELLER (8v1) run there is a summation of the > quality of the model:
I recommend you use 8v2 rather than 8v1, as this corrects some bugs.
>>>Summary of successfully produced models: > > Filename molpdf > ---------------------------------------- > A.B99990001.pdb 2355.08740 > N.B99990002.pdb 2376.38696 > ... > ... > > My question now is if these values really are the molecular pdfs or if it is the > objective function?
The two terms are interchangeable in this context.
> Another question: What is the default level of refinement actually? I have not > been able to deduce this from the manual.
automodel is all Python code, so you can find out by looking at the code (in modlib/modeller/automodel/automodel.py) or simply by printing out the current value:
from modeller.automodel import * env = environ() a = automodel(env, alnfile = 'alignment.ali', knowns = '5fd1', sequence = '1fdx') print a.md_level
This will print something like: <function very_fast at 0x400794c4>
i.e. the default level of refinement is very_fast.
By looking up the very_fast function in modlib/modeller/automodel/refine.py you can see the exact annealing schedule used.
Ben Webb, Modeller Caretaker
Citat Modeller Caretaker modeller-care@salilab.org:
> Leyla Celik wrote: > > In the end of the log-file from a MODELLER (8v1) run there is a summation > of the > > quality of the model: > > I recommend you use 8v2 rather than 8v1, as this corrects some bugs.
Yes, but 8v2 was not released when we built our initial models and we would like to be able to compare the works directly.
> > >>>Summary of successfully produced models: > > > > Filename molpdf > > ---------------------------------------- > > A.B99990001.pdb 2355.08740 > > N.B99990002.pdb 2376.38696 > > ... > > ... > > > > My question now is if these values really are the molecular pdfs or if it > is the > > objective function? > > The two terms are interchangeable in this context.
But as far as I understand the objective function is -ln(mol pdf)... I have two models from the same alignment run in modeller at two different times, one with a value of ~2000, the other at ~8000. If these represent the objective function the first model would be the superior whereas it would be the opposite if it is the molecular pdf. As I see it, there is quite a difference here!
Leyla
> > Ben Webb, Modeller Caretaker > -- > modeller-care@salilab.org http://www.salilab.org/modeller/ > Modeller mail list: http://salilab.org/mailman/listinfo/modeller_usage >
participants (2)
-
Leyla Celik -
Modeller Caretaker