Example of predicted functional sites colored by prediction confidence.
A protein structure can often reveal the details and mechanism of its function. Just over 50,000 protein structures have been experimentally determined but with over six million unique proteins sequences deposited in various databases there could be quite a wait before your protein is solved.
In the meantime, what can you do? Firstly, you can nominate your protein to be solved by the PSI. Secondly, you can generate a model of your protein based on its predicted structure. This should reveal surface residues and conserved amino acids that are likely to form the active site or binding pocket.
Numerous protein structure prediction programs exist, but one of the easiest to use is the Phyre server 1 . The predictive power of this server is similar to other publicly available software, and the most recent Critical Assessment of Structure Prediction (CASP) competition judged the latest Phyre version to be one of the best servers for preliminary assessment of protein structures.
Getting started is simple: go to the Phyre home page (www.sbg.bio.ic.ac.uk/phyre/), paste in your amino-acid sequence, add your email address and click the button. Wait half an hour, and you will receive an email with a link to a web page with the results, which include downloadable three-dimensional models of the protein and an estimate of how much confidence can be placed in these predictions. The coordinates of the model are in PDB format, so they can be viewed by standard molecular viewers such as Pymol, Rasmol and others.
The models generated rely on homology modelling, which is the most accurate way to predict protein structure. An estimated precision is given for each structure and the confidence values are color-coded from red (high confidence) to blue (low confidence). If several choices are presented, choose the one with several high-confidence predictions.
Now that you have a reasonably accurate model, how do you predict the residues that are most likely to be involved in the function of the protein? Click on the 'Model pockets/cavities/clefs' button and those residues found within the five largest pockets will be labeled. More information can be found in Nature Protocols.
So what are your options if Phyre or other databases can't find a good homology match for your protein? You could try an ab initio structure prediction approach such as Robetta. Alternatively try Phyre again in a few weeks' time. With 50 new structures solved each week, a reliable homolog may well be there next time you check.