I would like to generate a fragment library based on a sequence. I tried running the fragment_picker demos in rosetta3.3/rosetta_demos/fragment_picker and it seems I need to include a pdb and also psipred (.ss2) file. The documentation seems to say that only vall .fasta .wghts are mandatory. I only have sequence -- can I still run this protocol?
I think if I don't have any third party secondary structure information or a pdb I somehow need to change the .wghts file. For instance the .wghts given in the demo reference predA (BestFragmentsProtocol/input_files/simple.wghts) or psipred, sam, jufo (QuotaProtocol/input_files/quota-protocol.wghts).
A) You can try changing the weights to zero in the weights files you mention. There should be a real-number-valued column to change to zero (or try deleting those rows). I don't know which weight to leave on for sequence-only prediction; play around or maybe it's in the documentation?
B) You can create dummy secondary structure prediction (all loop, if you prefer) and run with that?
> it seems I need to include a pdb and also psipred (.ss2) file.
You need to provide the query amino acid sequence. It may come in the form of FASTA file, PsiBlast checkpoint or a PDB file. psipred file is not required. Assuming that you have only the FASTA file, your weights file would look as follows:
RamaScore 150 2.0 -
SequenceSimilarity 200 2.0 -
and your flag file:
-frags:frag_sizes 9 3
Note, that without secondary structure prediction, you can't use the quota protocol
RamaScore 150 2.0 -Robetta (which isn't as flexible as fragment_picker) generates the .jufo_ss, .psipred, and .psipred_ss2 files which can be used this fragment_picker protocol