Maybe u have to use Crystallography. You cannot predict for 100% without Structure Analysis.

In my opinion there is no way for 100% 3D Structure without practical work.

Dear Hadis,

I have posted on other threads my views on the protein structure in general. Recasting the general views onto what you asked for in here, there are several points to be made:

1) Sequence is only loosely associated with the structure (change of paradigm) so I have seen up to 25% sequence identity and radically different structure. But even the same sequence can have many different structures (usually related).

2) Proteins have a solid-liquid state mixed characteristics so, only around 30% of proteins have well defined structure at most time and conditions.

3) Even if proteins have well defined structure they can have mobile parts that usually imprint function.

4) All theoretical tools have to fail in the rate proportional to the predictability of the structure belonging to the folding class.

5) The course of action you should take is you should go the the XtalPred server that tries to determine whether the protein is crystallizable. As a byproduct you are going to get the answer what elements of your structure are intrinsically disordered, what elements may form what kind of secondary structure and how many similar proteins with defined structures are present in the databases. The next, you should try to predict domains by any of a number of web services, and finally try to predict (from the sequence) only the secondary structure.

Once you have done this you would know so much more about your protein that you would be able to estimate whether at least you might expect something useful to be obtained from automated services. In the very end you may try some protein model validation servers that actually can estimate the quality of your model by consensus search through different services.

I hope it helps.

So in essence Felix is right.

Mr Boguslaw Stec

Thanks a lot for your exact and attractive answer.

It was a beneficent guidance.

Best regards

Try TSPpred, participated in recent CASP11 assessments.

http://crdd.osdd.net/raghava/tsppred/submitseq.html