[modeller_usage] Ab-initio loop modelling vs. homology loop modelling

19 Feb 2015


      Dear Modeller-Usage group,
I have a problem with my structure. It is lacking 9 amino acids from a loop
next to its active center. I've tried to refill it using the loopmodel
class script from this tutorial:
http://salilab.org/modeller/wiki/Missing%20residues. My script looks like
this:
# Ab-initio modeling by the loopmodel class
from modeller import *
from modeller.automodel import *    # Load the automodel class
log.verbose()
env = environ()
# directories for input atom files
env.io.atom_files_directory = ['.', '../atom_files']
a = loopmodel(env, alnfile = 'sEH_AN_A.ali',
              knowns = 'sEH_AN_A', sequence = 'sEH_AN_A_fill',
              assess_methods=(assess.DOPE,
                              assess.GA341),
              loop_assess_methods=(assess.DOPE,
                              assess.normalized_dope,
                              assess.GA341))
a.starting_model= 1
a.ending_model  = 10
a.loop.starting_model = 1
a.loop.ending_model   = 10
a.loop.md_level       = refine.fast
a.make()
I've also find a homologous structure with a complete amino acid sequence
but its sequence has only 34% identity with my structure. So I just cut the
loop from the homologous and use it as a template. I think that the loop
from my structure will be similar to the loop from the homologous
structure. My second script looks like that
# Homology modeling by the automodel class
from modeller import *
from modeller.automodel import *    # Load the automodel class
log.verbose()
env = environ()
# directories for input atom files
env.io.atom_files_directory =  ['.', '../atom_files']
a = automodel(env,
    alnfile  = 'sEH_AN-homolog.ali',
    knowns   = ('homologue', 'sEH_AN_A'),
    sequence = 'sEH_AN_A_fill',
    assess_methods=(assess.DOPE,
                              assess.normalized_dope,
                              assess.GA341))
a.starting_model= 1
a.ending_model  = 100
a.make()
Then I compared all the models I've build by Z-score (normalized.DOPE)
value and by calculating the RMSD of the created models and homologous
structure. The RMSD is very high (1,55 to even 5,48) so I think the models
are not so good.
Which method will give me more accurate and native-like models?
Best regards,
Karolina Markowska
Silesian University of Technology