Frank van de Loo (FJVAND00 at ukcc.uky.edu) wrote:
: This may be laughable, but I need to know. Can one construct a model protein
: structure from scratch? If one has a protein of a few hundred amino acids
: length and various types of data (homology, site-directed mutagenesis,
: photolabelling, etc), and can come up with a hypothetical super-secondary or
: tertiary structure on paper for most or part of it, is it possible to constuct
: the model in a program that can then display it and be used for further
: refinement? I have been trying to do this in Hyperchem, with much frustration
: since I can't fold the polypeptide about very well once I've made it. Is there
: a way, or am I just trying to do something impossible? All advice greatly
: appreciated!
The consensus among workers doing algorithmic protein folding seems to be
that it's very difficult to do better than about 65% correct predictions
using the primary sequence alone. Homology to known structures and other
evidence is very helpful to place constraints on the problem; a fairly
decent model of the HIV-1 protease was published before the crystal
structure was published, and the agreement was fairly good. So yes, it has
been done, and more than once. It requires a great deal of effort with most
software systems I've encountered, and a fanatic attention to detail. I'd
also suggest getting some NMR data, like inter-proton distances from NOE
experiments, and internal vs. surface residues from D:H amide exchange, if
at all possible, or at least some CD measurements. The value of
site-directed mutagenesis data is less clear, unless done in conjunction
with NMR or some other probe, since the mutation could change the structure
more than might be expected in some cases.
--
Rick Venable =====\ |=| "Eschew Obfuscation"
FDA/CBER Biophysics Lab |____/ |=|
Bethesda, MD U.S.A. | \ / |=| / Not an official statement \
rvenable at deimos.cber.nih.gov \/ |=| \ or position of the FDA. /
