Modeling Tertiary Structure
Predicting and modeling protein structure, making protein crystallography somewhat obsolete, has been an elusive science to date. Since Linus Pauling and others determined that the most favorable secondary structures are the alpha helix and beta sheet, molecular and structural biologists have sought a set of rules with which to predict a protein’s structure from its amino acid sequence.
Structure prediction is still more guesswork than anything however, even with the power of 21st century supercomputers. It is based on comparisons to previously solved structures, not on reductionist rules and laws of chemistry.
Except maybe not. A news bite in last Thursday’s issue of Nature, the shape of protein structures to come, passes the word that an audacious project has succeeded in predicting the structure of a protein using just its amino acid sequence. The project, Ewen Callaway writes, used millions of hours of computing time donated by users of 150,000 home computers, similar to the Seti@home project. The results are still not based on a set of laws, and are still based on comparisons with known structures and their sequences (just smaller sequences), but it does seem to be a step forward, they say. This Rosetta@home project took millions of computing hours to get from a 112-amino acid sequence to a series of candidate structures, and many proteins of interest are much larger. How many billions of home computers will be needed to use this program more broadly?
It seems to this writer that the complexity of protein structure, with seemingly countless degrees of freedom to consider for determining energetic optima, is just beyond the computing power that we have at our fingertips. Perhaps with innovation relating to quantum computing however, this program could be applied to more than just one or two proteins. Until then, keep BLASTing.
The results of this study were pubished online at Nature on October 14th:
B. Qian et al. Nature doi:10.1038/nature06249; 2007
molecular biology
