So what I need (also described in a link above) is that I have RNA pdb file with hundreds of residues in witch I miss usually about ten segments each max 40 but often 3-4 residues ( I have only FASTA sequence of the missing parts). I want to predict the missing parts and insert it into my pdb.
That should be the protocol to use, unless you want to look at the stepwise modeling protocol. The stepwise protocol is a bit more involved in stringing together, though, as it requires a bunch of file manipulations and organization on your part.
With either protocol, my suggestion would be to – at least for your first run – not try to fill in all 10 sections at once, but instead look at remodeling just a single segment, probably one in the 5-7 nt range. Once you’ve figured out the protocol for modeling a single missing segment and have results for that segment which are looking reasonable, only then would I suggest trying to move on to the more complicated protocol of filling the multiple other missing segments.