In an ab_initio run with restraints, is there a preferred way to introduce a hydrogen bond ie. in cases where a sheet is known from a series of reliable observations ? Is it acceptable to include an HN-O/N-O restraint or is there a better way that allows the Rosetta engine to handle the details of geometry/distance?
A) forcing sheet in your secondary structure predictions from which fragments are generated is a good idea
B) hydrogen to acceptor distance constraints along the sheet are probably the cleanest thing to use. I wouldn't bother with angle/torsion constraints - Rosetta should pick up the hbond and do that automatically. You can use AmbiguousConstraints if you aren't quite sure of the pairing - is that your case?
C) There is likely to be a much better way to force sheet pairing with some sort of external input file...Rocco, do you know?
Constraints are definitely the way to do it (after the fragment secondary structure enforcement). After a quick check, I don't see any constraint that is specific for sheet-sheet pairing. Atom pair constraints are probably the way to go.
The remodel protocol (which I believe should be in the 3.5 release) has a bunch of facilities for enforcing certain tertiary structure geometries, but that's set up primarily for protein design where a large number of interactions are being made, rather than ab initio folding with one particular pairing.