I am not 100% clear on your question, are you hoping to model by searching the entire PDB library? If that is your goal, your best bet is to run your sequence through something like Phyre 2, it will model based on the closest possible structure. I do this and then remodel based on the top 3 hits from Phyre, manually using modeller.

 Dear Micah,

thanks for your answer! this is a good method.  But, when we don't have good template, we want to screening with all the protein in the PDB as template.

sincerely

Liuqing Chen

Hi, homology modelling by modeller can create protein model that has as low as 30% sequence similarity.

If at all, there is no template, you can try QUARK on line server that model the protein based on ab initio criteria. This is freely available by Zhang lab, and can model from 20-200 amino acids. 

http://zhanglab.ccmb.med.umich.edu/QUARK/

Good luck.

Muhammad is correct in suggesting QUARK. However QUARK is best when searching active sites because it is limited to 200 AA. You didn't mention your sequence length or goals so it is hard to suggest alternatives. 

I understand you don't have a good template. Did you by chance try the PHYRE 2 service (free and online)? It tends to literally try any protein on the whole PDB database. I myself have used it to try and model a protein with unknown function which did not align to much through a BLASTp search. It wasn't a perfect output, but it gave me ideas to properly model with a few templates after that.

Thanks all for suggestions. 

In actually, i mean, if we only  have crystal diffraction data and don't have the sequence data, so we want to try all the protein in PDB database as search template.

I propose you redo your diffraction using SAD or MAD to get your phases. Then with the resultant electron density map, you build your model manually (COOT). It is Hard tedious work, but you won't have Molecular Replacement bias,  you will learn a lot about where every atom is in your model and you will probably have a very reliable model (depending on your resolution) that you can publish and sleep good at night.

Dear Wes

you're right!  it's hard to get the phase by MR in my situation.  I planed to do SAD.

sicerely

Liuqing chen

Dear Liu-Quing

We're actually working on a server to do exactly this, based on the experience reported in this paper

http://journals.iucr.org/d/issues/2016/08/00/nj5253/index.html

It will probably be called SIMBAD and will be hosted at CCP4 online.  Watch the CCP4 bulletin board for an announcement. If your case is urgent then you can contact us by e-mail.

Regards

Daniel Rigden

 Dear Daniel

In fact, i  asked this question after reading that paper, 😁.     After  i reading the paper,  i  think this method is very good,  so i  asked the question.

But i  don't  good at computer technology, don't know how to run scripts, especialy on linux.   So,  if there has a tutorial, it's will be very good.

Anyway,  thank you!

sincerely

Liuqing Chen