Take the 2-minute tour ×
Biology Stack Exchange is a question and answer site for biology researchers, academics, and students. It's 100% free, no registration required.

I'm working with a protein that doesn't have any homology to other proteins so it will likely require ab initio structure prediction. However, since I don't work for a structure prediction lab and I don't have a stringent requirement to have a high resolution structure, I'm fine with a 6-10 angstrom resolution prediction to help me visualize the protein.

I'm imagining I-TASSER to get an initial idea of the topology. SWISS-MODEL doesn't find any homologs to template against. I'm trying to avoid using Robetta since it will probably take a month. What other low-resolution webservers should I test?

Alternatively, you could just tell me to suck it up and submit it to a month-long prediction.

share|improve this question
add comment

2 Answers

up vote 5 down vote accepted

Have you used hhpred to search for homologues? What was your criteria for defining there was no homologues? You could potentially go down to around 30% sequence identity to model.

I would submit the the query sequence to both I-Tasser and Rosetta and see if both of the servers agree on the topology. I-Tasser will always provide 5 models of your query sequence (even if they are rubbish) ranked by a confidence score.

While you wait for the above servers to finish, I would submit the sequence to a number of secondary structure prediction servers to come to a general consensus for the secondary structure of your protein. Some servers you could use are:

If the servers come to a consensus, compare the secondary structure prediction with that of the secondary strucutre of the models, to see if the ab initio modelling concurs with that of the secondary strucutre prediction. This will give you more reason to be confident in your model.

I would then validate your models using:

  • Prosa - This calculates a Z-score, and plots where it fits with structures with a similar amino acid length, from the PDB.
  • DOPE Score - This is employed in Modeller and measures the "nativeness" of your model, the lower the score the better. But you will need to install Python and have some knowledge of how to use the command line. I can provide the code to Run Dope with modeller.

Finally, are there any constraints you could introduce such a oligomerisation site or a protein-protein interaction site, to help with the modelling?

share|improve this answer
add comment

Robetta's the best service I've found for a protein I'm working on that doesn't have a solved structure. It's down right now for the CASP10 competition, but I would submit it and wait. The wait time lately has been ~2 months.

share|improve this answer
    
CASP10 would certainly explain the wait time. Guess I'll just add it in and download Rosetta@Home to make it go faster (slightly). –  bobthejoe Apr 13 '12 at 23:33
add comment

Your Answer

 
discard

By posting your answer, you agree to the privacy policy and terms of service.

Not the answer you're looking for? Browse other questions tagged or ask your own question.