Protein folding problem solved?
rvenable at deimos.cber.nih.gov
Fri Jun 9 23:34:11 EST 1995
On 9 Jun 1995 11:15:57 -0400 Clifford Beall pontificated:
> In the June issue of the Johns Hopkins Alumni magazine (I only quote from
> the most authoritative scientific sources), it is claimed that the protein
> folding problem has been solved by two Hopkins researchers: Rajgopal
> Srinivasan and George Rose. They have developed a computer program known
> as LINUS that it is claimed can predict the folding of soluble globular
> proteins. The pictures that appear with the article are very impressive.
> It seems like this could be a massive discovery. Would anyone with
> knowledge of this work like to comment further?
Several researchers have made this claim over the last few years, including
Harold Scheraga, Peter Wolynes, and (as someone has already commented) Ken
Dill. I'd call it hyperbole; in most instances these methods can attain
better success rates than classical Chou-Fasman or Garnier-Robson-Osguthorpe
sequence predictions, but that doesn't mean the problem is solved. In
general most of this kind of work is limited to structures in the Brookhaven
PDB; that represents a subset of smaller, water soluble, freely diffusing
proteins, or else minor fractions of a much larger protein (e.g. the
cytoplasmic receptor or an FAB fragment). Given that the crystal is a
somewhat artificial environment, and the large number of proteins whose
structure has not been determined, and/or are not included in this type of
work, I would greet any such reports with a great deal of skepticism. I
would challenge anyone claiming to have "solved the protein folding problem"
to predict the structure of the HIV envelope protein gp160, or perhaps the
complete acetyl choline receptor before making any such global statements.
There are actually several promising approaches to predicting the structures
of soluble, globular proteins, including genetic algorithms and neural nets.
Dealing with this minor subset is a natural and logical starting point, but
is hardly representative of proteins in general. It is interesting work
overall, but tends to be oversold; I'd love to see a competition with each
structure prediction group given the same set sequences chosen by an impartial
panel under controlled conditions.
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. /
More information about the Molmodel