Abstract

Protein structure prediction: what works, what doesn't and what use is it?

Center for Advanced Research in Biotechnology University of Maryland 9600 Gudelsky Drive Rockville, MD 20850, USA

tunc@indigo5.carb.nist.gov

Computational methods for determining protein structure from amino acid sequence have apparently been advancing rapidly. But just what can they currently deliver? The results of a community wide experiment made to determine the state of the art will be described. The experiment consisted of three stages: collect information about soon to be solved structures and inform the prediction community, collect predictions before the correct answers were public, and analyze and discuss the results.

Predictions were divided into comparative modeling, where there is a clear sequence relationship to a known structure; threading, where a new sequence is tested for compatibility with each type of known fold, and traditional ab initio methods that do not rely directly on databases.

Results in each area will be summerized, with particular emphasis on the usefulness of the methods for guiding experimental design and for the identification of protein function from sequence. Prospects for the future will be discussed.

References:
PROTEINS vol 23, no. 3, 1995. (This issue is devoted to papers describing the outcome of the experiment)
J. Moult 'The Current State of the Art in Protein Structure Prediction' Cur.Opin.Biotech. vol 7 422-427 1995.