[modeller_usage] working & accuracy of modeller

Dear all,

I have some basic queries regarding modeller working & its performance.

1) What is the accuracy of model building by modeller?

2) I have read in the modeller mailing list that in the case of multiple
templates, modeller uses weighted sum over all the templates. Does this mean
that spatial restraints are derived individually from the alignment between
first template & query sequence and from the alignment between the second
template & the query sequence? Then these resraints are added and weights are
applied to that template according to the local similarity with the query
sequence? Anyone please correct me if I am wrong?

3) I have built a model using 2 templates with modeller9v2. RMSD (only backbone
atoms) between the model & the first template is 1.6 ang & with second is 3.3.
Sequence identity is higher between the target sequence & first template as
compared to that of the second template. How can I calculate the accuracy of
this model in terms of percentage? What is the way of calculating the accuracy
of model building?

Looking forward to your reply? Kindly reply and enhance my knowledge.

Thanks in advance

Ruchi