[modeller_usage] DOPE and sequence/model length

[Starr Hazard <>]

Folks,

I have modeled some dimers of a large protein and a much shorter partner 
(~350aa for X and about 80aa for Y) . There are crystal templates for 
such dimers. The best resolution structures have  different large and 
small proteins. So I have templates with  X'Y' and X''Y''.  X' and X''  
are homologs as are Y' and Y''.  The Y family is quite variable in 
length. When homologs from distant organisms ( eg plant, fungi, fish 
mammal) are aligned there are gaps compared to the mammalian templates.

In mammals X' and X'' are in related but distinct clades. If I model a 
yeast X protein to each mammalian template can I say that the yeast 
protein is better represented by one or the other template based on a 
DOPE score? ( the GA341 scores are not helpful since they are always 
1.000 for these models)

This is several questions I guess.  1) Are DOPE energies normally 
distributed?
                                                               Could I 
statistically compare DOPE scores of models made with one template to 
those from another template.
                                                           2)  I have 
some evidence that longer XY pairs have "better" DOPE scores than 
shorter pairs. Is there a statistical way to control for "length"?
                                                           3) I was 
just looking at DOPE scores for two models. One with from an alignment 
with gaps and the other from an alignment without  gaps.
                                                                The PDB 
model files are different sizes yet the scores are identical. Does DOPE 
ignore regions that are not in the template?

Starr