Despite decades of work, understanding how proteins fold remains a major research challenge. The fruits of this massive research effort have been: development of (i) methods for predicting the likely structures that protein sequences will adopt, or for simulating the folding process itself; and (ii) databases of structural information (e.g., containing 3D coordinates, fold classifications, structure summary data, and so on). As part of the ongoing endeavour to understand the principles of protein folding, we have been involved in the development of a new, integrated structure information resource, based on a small subset of the PDB . The resource contains information derived from a combination of sequence analysis tools, structure analysis software and fold simulation algorithms; to make the contents more accessible to the wider community, we have also developed a user-friendly front-end for visualising the integrated data. The motivation for combining data from these various approaches is to offer insights into the role of particular types of residues and fragments in protein folding, and hence to improve our understanding of factors that are critical to the folding process in general.
A structural annotated database has been generated derived from several unrelated algorithms and data sources for an integrated analysis of critical fragments in protein folding. From an initial analysis of the data, we found, not surprisingly, that certain results were strongly correlated: e.g., residue accessibility values (denoting the degree of internal constraint on flexibility), Fold-X  scores (denoting the stabilising contributions to the fold), Popmusic  values (denoting destabilising contributions), and lattice simulations  (denoting the number of close neighbours or interaction partners within the fold). We used these values to synthesise a 'folding score'.
The PFF database collected analyses and their companion resources to be made publicly available. A goal of the PFF consortium was to create a consensus "prediction" tool combining the strengths of different methods. We found that integration of different methods has indeed added value over individual ones. Coupled with the degree of conservation of residues, a folding score was created to delineate regions that are likely to contribute to (i) the stability of the fold (and hence may contribute to the folding nucleus), and (ii) the function of the protein. This offers a means of automatic motif detection, which can be used for protein family characterisation and functional/structural annotation of evolutionarily conserved regions. We present here a simple case-study to illustrate how the combined data can be used to pinpoint such motifs with potential structural and functional roles.
Version 1.0 of the PFF dataset is accessible in a DSSP-flat-file format from http://babylone.ulb.ac.be/LIFE/ webcite; it is also available in an XML format through the UTOPIA toolkit, together with the UTOPIA visualisation tools for OS X, Windows and Linux at http://utopia.cs.manchester.ac.uk webcite. The Web resource for calculating combined folding scores is accessible at http://umber.sbs.man.ac.uk/~corpas/db/ webcite. For a more detailed explanation on the meaning and biological implications of the folding score please refer to http://umber.sbs.man.ac.uk/corpas/db/method_doc.html webcite.