building a 77 aa long protein fragment

[Anonymous]

Hi guys,

The same question was posted in PyRosetta forum but since questions are answered faster here I’m reposting it.

I am trying to generate models of a 281 aa long protein domain with Modeller. The problem is the low sequence identity with the templates (ranges between 30-40%) and an unaligned region, which is 46 aa long, but induces a conformation disorder spanning from residue 84 to 161 (77 aa long). So I though to build this fragment of the model with Rosetta and optimize the structure. My questions:

1. I’m new to Rosetta, hence I’m not sure if what I am after is possible. (?)
2. If yes could you point me to the appropriate tutorials/scripts?
3. Is it easier to use PyRosetta? for this task rather than Rosetta? (Unbiased opinions please)

I would greatly appreciate any advice.

Tom

[Anonymous]

Hi Tom,
I wanted to clarify your question before I attempted an answer. You are trying to build structural models of a protein with 281 amino acids. You have templates( homology models )for part of the protein, but not for an unaligned region( 46 amino acids ).

For residues 84-161 a conformational disorder is induced( is this a result of the alignment in Modeller? )

Ultimately you would like to build structural models of the whole 281 residues and you hope to use RosettaAbinitio to predict residues 84-161?

please give as much detail as possible about your system

Grant

[Anonymous]

Hi Grant,

thanks for getting back. I do have templates (crystal structures) which are aligned with all the target sequence apart from this 46 aa. long part. As a result, the predicted structures from Modeller have a spaghetti-like conformation which extends from residue 84 to 161 (this region includes the unaligned part). So I was think if it is possible to feed a Modeller-generated structure to Rosetta and tune it to repack residues 84-161? If not then I could predict the whole structure ab-initio and use the Rosetta-predicted structure as a template for Modeller (by tuning it to read only the 77 disordered amino acids to avoid bias towards the Rosetta-generated structure – obviously because the sequence identity between the latter and the target sequence is 100%). Any other suggestions are welcome.

cheers,
Tom

[Anonymous]

Hello Tom,
It sounds like what you want to do is to do RosettaAbinitio where only the structure of the 77 disordered amino acid portion is predicted. This is possible if you are willing to modify some code( only a few lines and I would give you good directions).

[Anonymous]

Theoretically will the predicted structure of this fragment be different when repacked than when building the whole structure ab initio? If not then I don’t find a reason to go through source code modifications. Otherwise I’m willing to do it, so please guide me.

[Anonymous]

Since this region is not conserved it could serve as an ideal target site for drug design, so please get back to me with details.

[Anonymous]

Hello Tom,
Sorry about the delay. I’ve looked into the code modifications and it would be a lot of work for someone not directly working on rosetta. If you chose to use rosetta to predict this 77 aa region, I would predict the whole structure and add distance constraints for the regions which you have homology models. Good luck.
Grant

[Anonymous]

Hi Grant,

I presume you mean to add constraints in Modeller when using the Rosetta-generated model as a template. But isn’t it possible to generate constraint files (i.e. NMR-like constrains) for the rest of the structure (excluding the disordered loop) from the homology models and use them in Rosetta? Are you aware of any script or program that can generate constraint files from a .pdb file?

In this way the ab initio structures will be more consistent with the homology models. Moreover the packing of the regions where the alignment is good may have an impact on the conformation of the loop, and since Modeller tends to be more reliable in prediction of evolutionary conserved regions, the ab inition predictions will be more accurate. Does this make sense?

cheers,
Tom

[Anonymous]

Hello Tom,
I was saying that you should fold the entire protein in Rosetta( all 281 amino acids ). And to improve the quality of your results you should set distance constraints between Calpha atoms that you know to be close to each other from your Homology models. The constraints are well documented.
Grant

[Author]

Posts