is it possible to ab initio only a part of a protein?

Why not just use CCD fragment-based loop modeling? It’s good with loops of that size.

It’s worth a shot. The fragments should bias it towards whatever secondary structure it is supposed to be. CCD does not solve to prefer loops, it’s just that it’s used for loops, but is not really smart enough to keep secondary structures intact. Hopefully your fragments will get inserted and nearly close the “loop” and won’t get modified much by the CCD part.

I have asked the abinitio people repeatedly to please document how to do “partial abinitio” as you request (you are not the first person to ask for it) and have never gotten a response. I’m sure it’s possible but I don’t know how. It may be as simple as playing with the movemap; if you want to try editing the C++ code directly we can try to see where the movemap is and tweak it to only allow your regions to change.

It’s not strange if you think about it. This level of development in Rosetta is done by grad students and postdocs. Our incentive is to publish papers, not documentation. We do our best, but it’s always challenging to get the code authors to write documentation on top of papers and code. It doesn’t help that by the time a protocol has been published, it’s several years old already, so the person who wrote it may have graduated or at least moved on to other projects.