Dear Rosetta experienced users,
I am trying to predict the structure of a protein using the abinitio algorithm.
The protein is found in nature (not synthetic), but its structure has not been solved yet, and it proved difficult to crystallise. The psipred secondary structure prediction shows that it has 2 helices and some strands, but it also has large sections of loops.
The protein is 127 amino acids, and there is no crystal structure, therefore I have no reference structure for an RMSDvsSCORE plot.
I tried to predict the structure by generating around 500,000 structures using abinitio, but it did not give a good prediction. The clustered 200 lowest scoring decoys are not the same.
My questions are:
1. Is the prediction still possible with a large percentage of loops? Would loops be floppy, give intrinsically disordered segments, and hence give a bad prediction?
2. Is the difficulty of crystallising the protein an indication that it is also difficult to predict with abinitio?
3. Did I run the prediction incorrectly? Are there certain parameters for predicting a protein without a reference structure? My protein has several Cystines, could this be an issue? What is the best practice to run such a protocol?
4. Can someone just talk to me about the subject of using Rosetta to predict the structures of yet unsolved proteins so I can find a better path to teach myself this field?