Hi
I have a sequence for which I could not find a suitable template to model. Is there any other method to generate a 3D structure?
I submitted to 3D-PSSM but could not get an appropriate template. So can you tell me how to find a suitable template?
The short answer is that if your protein sequence lacks sufficient similarity to a protein with a known structure, then comparative modeling methods are not going to work. And if you use an unsuitable template, any modeled structure that you do obtain is likely to be quite meaningless.
You might want to look at Rosetta, a highly-regarded suite of tools for protein structure prediction. However, be aware that there are many occasions when modeling simply won't work, due to insufficient information.
It's possible to use comparative modeling effectively down to 20-25% of similarity in alignment. The only secure way to proceed in the twilight zone is by threading your protein, domain by domain. It's a very manual process and time consuming. I've used this approach with some nuclear hormone receptor (TRs) and worked very well with excelent statistics in WHATCHECK and similars.
In order to proceed with threading, I suggest start with a search in Pfam and some predictions in PsiPred. This Biostar thread is very helpful too.
You can one of the automated Homology Modeling servers to generate a structure. For example: Try ModWeb or SWISSMODEL Automated mode. These tools can generate a homology model from a given fasta file using remote homologues, but consider models with reliable sequence similartiy for further analysis.
If homology to the single template is low. In the twilight zone of similarity, One approach you can follow is "multiple" template approach to cover the protein sequence with multiple overlapping regions from different structural templates.
search your sequence using PSI-blast with multiple iterations against pdb. select the best possible templates.
See this paper you will get a fair idea aboutthis approach.
3D-PSSM template database is not update any more (it's from 2004!) as stated on its website. It has been superseded by Phyre.
I would recommend to try HHPred, which is one of the best, if not the best template finding server.
To try other servers, the fastest way is to use a METASERVER (e.g. http://genesilico.pl/meta2/ from my previous lab), which will submit your sequence to several servers automatically.
If you indeed cannot find a template with those tools, try indeed Rosetta or I-TASSER.
Note that de novo modeling more often fails than works, so I would think twice before spending time on that (do you expect to answer you biological question with very uncertain de novo model?).
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
I edited the title to make this question more specific (and removed the all upper case - no need to shout!) Also, we don't need your email address; questions are answered and discussed right here.
I've removed the mail. We don't want robots/fakes/similar scanning the site.