Proteins Suitable For A Structure Prediction Project ?

Question

[curiousgeorge]

Hello there !

I am pursuing a project where I intend to (required to) predict the structure of a protein using homology modelling. I have searched the internet a lot but couldn't find a suitable candidate. Can anybody already having experience in the area shed some light on the proteins which will be suitable for this purpose, like for example whose structures are not known and are under 400 residues?

Regards,

Answer 1

[Neilfws]

The median length of proteins from most organisms is < 400 and of the millions of proteins sequenced, some tens of thousands have structures. So in fact most proteins fit your example criteria.

I would work a little harder on a selection strategy.

[curiousgeorge]

Thanks, it soothes me to know that there are tons and tons of candidates out there !

Answer 2

[K]

I guess you did BLAST your sequence (for instance in NCBI), and check if any hit was linked to the PDB database, right? That would be one way to find a candidate. A less efficient way to do a similar thing would be to type in the name of your protein and see if it can be found at the PDB... However, when you said you have searched the internet a lot, it likely already included these possibilities.

Alternatively, some modelling servers actually do find out templates for you -if available-. I believe I-tasser is somewhere in-between homology and ab-initio modelling: http://zhanglab.ccmb.med.umich.edu/I-TASSER/about.html ...and starts by looking for templates. It will be slow, though.

Also, the Robetta.org server, which performs Ab-initio modelling seems to look for templates in its initial step, the "Ginzu", by which functional domains are tried to be assigned. It will take you some days to have a result, however; sometimes weeks if you request a full model or the server load is too high.

Hope this helps. Cheers,

K

Answer 3

[João Rodrigues]

Adding to Neilfws' answer:

I would look at the Protein Structure Initiative websites and their targets. Another interesting idea is to look for all structures deposited in the last half-year in the RCSB Protein DataBank and see which one pleases you from a biological point of view, and then try to extrapolate it for another organism. For example, you find hen lysozime and you try and model human lysozime. Don't try to find an extremely tough target. From what I understand, you are only required to model something, probably to learn the whole process and how it works, so the final result is irrelevant. Choose something appealing and easy to present, but that also has a few challenges (identity < 60% and a few gaps for example).

Crazy suggestions: E2/E3 proteins from the ubiquitination pathway.

[curiousgeorge]

Thanks K and Joao for very useful answers. I had not done the BLAST as K suggested but directly tried finding the proteins I had read in the literature in PDB. That obviously was not efficient but came across a family of proteins which are not in PDB. I am trying to explore if homology modelling will be feasible. I will provide info on them here once I am sure they serve the purpose.