There are tons of servers to predict protein structure, and some are
extremely accurate, depending on your particular protein. In general,
you can obtain experimental (x-ray diffraction/NMR spectroscopy) level
accuracy for proteins with a high similarity (>50%) to a protein with
experimentally determined structure. (This typically is about 0.5-1
angstroem root mean square deviation (RMSD) on average, which BTW is
lower than the average RMSD for about 300 IDENTICAL proteins whose
structures were solved by/between x-ray diffraction and NMR
spectroscopy.)
These servers and methods are evaluated at the biennial Critical
Assessment of Structure Prediction (CASP) experiments which give you
an idea of which servers and algorithms work well and when. There are
even some cases where, for small proteins and/or distant homologues,
you can approach that level of accuracy.
See the CASP web site for http://predictioncenter.org for the servers,
their rankings, etc.
It's mostly an issue of what you want to use the resulting structures
for.
--Ram
Ram Samudrala, Ph.D.
Associate Professor, Computational Biology
University of Washington
Seattle, WA 98195, USA
V: 1-206-732-6122
F: 1-206-732-6055
W: http://compbio.washington.edu
Sander Groffen <sander.groffen from cncr.vu.nl> wrote:
> Are you all talking about the same type of structure?
> If you want the tertiary structure (3D) there is no such thing, as DK
> pointed out.
> If you want the secondary structure, protein prediction servers might
> get you started as Dr Buxbaum pointed out.
> If you have no idea what kind of protein you have, a mass spectrometer
> can help to identify it.
> (If you have no idea what the difference between primary, secondary and
> tertiary structure is, go read a textbook on biochemistry).
> Sander
