R.R. Copley, C.D. Livingstone, R.B.Russell and G.J. Barton
Laboratory of Molecular Biophysics, The Rex Richards Building, South
Parks Road, Oxford. OX1 3QU United Kingdom. Tel: 44 865 275368
E-mail: cdl@bioch.ox.ac.uk rrc@bioch.ox.ac.uk
r_russell@europa.lif.icnet.uk, gjb@bioch.ox.ac.uk, WWW URL: http://geoff.biop.ox.ac.uk
Prediction of the folds of Synaptotagmin and the Beta subunit of Urease
The aim of our work is to identify the likely fold of a protein given a
family of sequences. This is achieved in two steps: Secondary
structure prediction (including prediction of buried residues) and
Secondary Structure Mapping (identification of the fold). A secondary
structure prediction is obtained from an accurate multiple sequence
alignment for the family by identifying positions of
insertions/deletions, conserved residues and hydrophobic conservation
patterns, and combining this analysis with the results of a variety of
conventional single sequence secondary structure prediction
algorithms. This strategy has previously been used to generate
accurate secondary structure predictions for the annexins, SH2 domains
and protein-tyrosine phosphatases prior to the experimental
determination of the structures [Barton et al.,1991; Russell et al.,
1992; Livingstone & Barton, 1994]. Specifically, helices and strands
were predicted by the methods of Lim, Chou and Fasman and Robson as
implemented in the ``Leeds'' prediction suite [Lim, 1974a; Lim, 1974b;
Chou & Fasman, 1978; Garnier et al., 1978; Eliopoulos, 1989].
Predictions of turn were made using the algorithms of Rose and Wilmot
and Thornton [Rose, 1978; Wilmot & Thornton, 1988].
Regions of the multiple alignment where insertions and deletions are
seen, or which are varied in composition across the aligned family of
sequences were assigned to coil (non-core secondary structure). The
intervening regions were assigned either to helix or strand by
referring both to the results of the classical prediction algorithms
and to patterns of residue conservation identified with the aid of the
AMAS program [Livingstone & Barton,1993]. The combined procedure
yields a final three-state prediction. Secondary structure predictions
by the method of Zvelebil et al. [1987] were used as a "casting vote"
where ambiguity between strand and helix assignment could not otherwise
be resolved.
For each prediction, a non-redundant set of sequences was extracted
from the current version of the PIR [Sidman et al., 1988] and Entrez
[Ostell,1992] databases. Multiple sequence alignments were made using
the AMPS package [Barton & Sternberg, 1987, Barton,1990]. Poorly
conserved regions within the resulting alignments were adjusted by eye
in the locality of gaps.
Given the secondary structure prediction, prediction of buried residues
and any additional knowledge, for example the location of catalytic
residues, we searched a non-redundant set of protein domains to find
folds consistent with the secondary structure prediction and other
constraining data. This was achieved by the application of a newly
developed technique known as Secondary Structure Mapping (SSM) [Russell
et al., 1994]. SSM finds topologies consistent with secondary
structure assignments and that are consistent with any experimental
restraints (i.e. disulphide bonds, active site residues, etc.)
available for a protein/protein family. First, all possible alignments
of secondary structure elements are generated between a query (i.e.
prediction) and a database (i.e. PDB) structure. These alignments (or
maps) are then filtered to remove those structures which are 1)
uncompact; 2) having ends of secondary structures too far apart to be
bridged by predicted loop lengths; 3) having poor beta sheet hydrogen
bonding; 4) lacking essential secondary structures (e.g. those with
active site residues); 5) do not satisfy topological constraints (e.g.
sequential beta strands are most often antiparallel); and 6) those
which are unable to satisfy any experimentally derived distance
restraints. The program is fast (able to search a representative set
of PDB domains in about five minutes) and is able to go from a large
number of initial maps (ie. prior to filtering) to a few structurally
sensible maps.
Proteins Predicted: Synaptotagmin C2 domain, Urease Beta subunit, The
mystery protein (!)
Barton, G. J.(1990). Methods Enzymol. 183, 403-428.
Barton, G. J., Newman, R. H., Freemont, P. F.& Crumpton, M. J. (1991).
Eur. J. Biochem. 198, 749-760.
Barton, G. J. & Sternberg, M. J. E. (1987).J. Mol. Biol. 198, 327-337.
Chou, P. Y. & Fasman, G. D. (1978). Adv. Enzymol. Relat. Areas Mol.
Biol.
Eliopoulos, E. (1989). The leeds protein secondary structure suite.
Garnier, J., Osguthorpe, D. J. & Robson,B. (1978). J. Mol. Biol.
Livingstone, C. D. & Barton, G. J. (1993). Comput. Appl. Biosci.
9, 745-756.
Livingstone, C. D. & Barton, G. J. (1994). Int. J. Pept. Protein
Res.
Lim, V. (1974a). J. Mol. Biol. 88, 857-872.
Lim, V.(1974b). J. Mol. Biol. 88, 873-894.
Ostell,J. (1992). Entrez sequences graphical user interface.
Rose, G. D. (1978).Nature, 272,586-591.
Russell, R. B., Breed, J. & Barton, G. J. (1992). FEBSLett. 304, 15-20.
Russell, R., Copley, R. & Barton, G.(1994). A method for fold prediction.
The method is still under development.
Sidman, K., George, D., Barker, W. & Hunt, L. (1988).Nucleic Acids Res.
Wilmot, A.C. M. & Thornton, J. M. (1988). J. Mol. Biol. 203, 221-232.
Zvelebil, M. J. J. M., Barton, G. J., Taylor, W. R. & Sternberg, M. J. E. (1987). J. Mol. Biol. 195, 957-961.
Structural Biology home page
Asilomar Conference home page
LLNL Disclaimer
Web page maintained by BBRP Webmaster (BBRPWebmaster@humpty.llnl.gov).
CONF-941241
Last modified on 1-11-95