From Protein Structure to Function with Bioinformatics.pdf

52 L.A. KelleyIt is unclear to what extent improvements in structure prediction have been theresult of increases in database size versus algorithm improvement. Although thesequence database has been growing in size exponentially, its information contenthas not. The vast majority of sequences added to the sequence database each yearare highly similar to those already contained within the database. Recent work(Chubb, Kelley and Sternberg, manuscript in preparation) illustrates that despite agrowing sequence database, homology detection with standard tools (such as Psi-BLAST) is plateauing. It is thus hard to imagine further significant increases inhomology detection based on mining the sequence database alone. It is also unclearhow much the recent improvements in ab initio structure prediction may be due tothe increased structural database containing ever better fragments of structure foruse in fragment assembly methods (Zhang and Skolnick 2005).The database of sequences and structure will continue to increase. Even if algorithmicdevelopment on structure prediction were to cease today, structure prediction accuracywould continue to improve. Ignoring protein design issues, structure prediction isan exercise in practical reduction of time and cost in determining protein structure.The desire to ‘solve’ the protein folding problem is alive and well as one of theholy grails of molecular biology. Yet even in the absence of such a ‘solution’ itseems likely that we will have usefully accurate models of most, if not all proteinsfound in nature within a reasonable time. Whether that time is 5, 10 or 50 years willbe down to the ingenuity and diligent work of both experimentalists solving structuresand genomes, and modellers mining the information therein for all it is worth.This is no longer a question of ‘if’ but ‘when’.Protein design however remains a very different and daunting task, founded ona deep understanding of protein folding. To understand protein folding is to understandhow the ‘software’ of DNA becomes the ‘hardware’ of functional proteins.It is to understand, at a fundamental level, the nature of living things. However,there may be no elegant solution to the protein folding problem. Nature does notnecessarily find an elegant solution; simply one that works. Reluctantly, we mayhave to be satisfied with a complex predictive framework. Nevertheless the hope fora simple, computationally tractable and hitherto undiscovered explanation for proteinfolding remains strong.Acknowledgements I would like to thank Dr. Benjamin Jefferys for his extensive help with theillustrations in this chapter.ReferencesAltschul SF, Madden TL, Schäffer AA, et al. (1997) Gapped BLAST and PSI-BLAST: a newgeneration of protein database search programs. Nucleic Acids Res. 25:3389–3402Bateman A and Finn RD (2007) SCOOP: a simple method for identification of novel proteinsuperfamily relationships. Bioinformatics 23:809–814Bennett-Lovsey RM, Herbert AD, Sternberg MJ, et al. (2008) Exploring the extremes of sequence/structure space with ensemble fold recognition in the program Phyre. Proteins. 70:611–625