BIOINFORMATICS<-->STRUCTURE
Jerusalem, Israel, November 17-21, 1996

Abstract


Assessment of protein fold predictions from sequence information: predicted alpha/beta-doubly wound fold of the von Willebrand factor type domain is similar to its crystal structure

Yvonne J. K. Edwards* and Stephen J. Perkins

Department of Biochemistry, Royal Free Hospital School of Medicine, Rowland Hill Street, London, NW3 2PF, UK.
*Current address: UK HGMP Resource Centre, Hinxton, Cambridge, CB10 1SB, UK.

yjedward@hgmp.mrc.ac.uk


The von Willebrand factor type A (vWF-A) domain superfamily was predicted to be structurally similar to the fold of GTP binding proteins using an integrated computational approach [1,2]. At the time of the analysis, no member of the vWF-A domain superfamily had been experimentally solved and hence the fold was predicted from sequence. The fold prediction was based on sequence data, i.e., the joint use of Fourier transform infra-red spectroscopy data on recombinant vWF-A, averaged secondary structure predictions, solvent accessibility predictions and automatic protein fold recognition techniques [3], together with validation from biochemical experiments such as site directed mutagenesis. It is interesting to note that there is no significant sequence or functional similarity between vWF-A and ras-p21. Following this prediction, a crystal structure of the vWF-A domain in complement receptor type 3 (CR3) was determined [4]. We have shown the crystal structure of CR3 is similar to the predicted fold for the vWF-A superamily. Structural features predicted for vWF-A, such as, the Mg2+ binding site, the number, location, arrangement and solvent accessibility profiles of secondary structures, were largely accurate [5]. Additionally, the automatic fold recognition program THREADER [3] produced variable results for individual sequences of the vWF-A domain superfamily, but processing THREADER results across the superfamily was shown to significantly improve the outcome of the analysis. We conclude protein structures can now be predicted from sequence data with reasonable accuracy using available structure predictive algorithms in an integrated fashion, and that the predictive accuracy in this case was principally limited at the periphery of the fold [5]. This integrated approach for predicting folds from sequence is being applied to other proteins such as the proteoglycan tandem repeat domain [6].

[1] Perkins S.J. et al., (1994) J. Mol. Biol., 238:104-119.
[2] Edwards Y.J.K. & Perkins S.J. (1995) FEBS Letters, 358:283-286.
[3] Jones D.T. et al., (1992) Nature, 358:86-89.
[4] Lee J.O. et al., (1995) Cell, 8:631-638.
[5] Edwards Y.J.K. & Perkins S.J. (1996) J. Mol. Biol., 260:277-285.
[6] Brissett N.C & Perkins S.J. (1996) FEBS Letters, 388:211-216.


Back to the Abstract Index.