My question regards the theory behind and how Modeller creates the
models. The case is that I have been using the Missing residue-script
from the Modeller wiki-page to fill in a gap of 14 missing residues. But
how does Modeller build the gap when it does not have any template
structure as in my case?
For such regions, the initial atomic positions are constructed using
internal coordinates; Modeller knows what the ideal bond lengths and
angles are for all of the standard residue types. Modeller cannot know
the orientation of each residue, however, so it simply builds an
extended chain (look at the .ini file). It then optimizes the structure
using a combination of the CHARMM forcefield (e.g. to minimize steric
contacts) and its own library of statistical protein preferences (e.g.
Ramachandran plot for phi/psi, and chi1-4 distributions).