D.T. Jones, R. Miller & J.M. Thornton
Biomolecular Structure and Modelling Unit, Dept. of Biochemistry and Molecular Biology, University College, Gower Street, London WC1E 6BT, United Kingdom

The listed predictions were carried out using an updated version of the method of Jones et al. (1992). A dynamic programming algorithm ('double' dynamic programming) was used to align the given sequence with the 'real' coordinates of each structure in a library of folds, taking into account residue pairwise interactions. Matching pairwise interactions is related to the requirements of structure comparison methods. The interaction environment of a residue i is defined as being the sum of all pairwise interactions involving i and all other residues j/i. This is a similar definition to that of a residue's structural environment, as described by Taylor and Orengo (1989). In the simplest case, the structural environment of a residue i is defined as the set of all inter-Ca distances between residue i and all other residues j/i. Taylor and Orengo developed a novel dynamic programming algorithm for the comparison of residue structural environments, and it is a derivation of this method that we therefore use for the effective comparison of residue interaction environments.

At the heart of the evaluation function used here is a set of pairwise potentials of mean force, determined by a statistical analysis of highly resolved protein X-ray crystal structures and the application of the inverse Boltzmann equation. In addition to the pairwise potentials, a solvation potential is also used. This potential is determined in a similar fashion to the pairwise potentials, except that the variable in this case is relative solvent accessibility rather than interatomic distance.

A modified set of potentials was used for all predictions apart from the first 3. The main difference between the old and new pairwise terms is a correction for the size of the proteins used to generate the potentials and the protein being predicted (D. Jones, paper in preparation). This correction allows more of the pairwise interaction information to be extracted than by simply truncating the potentials at 10 . The new solvation potentials are based on just 5 unequal relative accessibility divisions, rather than 10 or 20 equal divisions.

Target	No. of library folds		Potentials

xylanase		244            	unscaled/10 solv. divisions
bhted			253            	unscaled/10 solv. divisions
smanucecs		253            	unscaled/10 solv. divisions
bphc			253            	scaled/5 solv. divisions
ce-1			253            	scaled/5 solv. divisions
urease			253            	scaled/5 solv. divisions
l14			266            	scaled/5 solv. divisions
pbdg			266            	scaled/5 solv. divisions
ppdk			266            	scaled/5 solv. divisions
rtp			266            	scaled/5 solv. divisions
staufen			266            	scaled/5 solv. divisions
synapto			266            	scaled/5 solv. divisions

References 1. Jones, D.T, Taylor, W.R. & Thornton, J.M. (1992) Nature 358, 86-89.
2. Taylor, W.R. & Orengo, C.A. (1989) J. Mol. Biol. 208, 1-22.

Asilomar Conference home page