Traditional protein biochemistry and classical purification was instrumental in identify­ing key wall biosynthetic enzymes, for example, XG:1-2,fucosyltransferse (116, 117) and GM:1-6,galactosyltransferase (88). As most of the glycosyltransferases involved in polysac­charide synthesis are membrane bound and are of low abundance, i. e., estimated 2-6 molecules per cell, it is clear that other methods should be sought to identify wall biosyn­thetic proteins. Classical genetics has already contributed immensely to the identification of large numbers of genes involved in glycan synthesis such as cellulose synthase and their large gene family (507,508) andxyloglucan synthesis (120). The genome sequence facilitated the identification of mannan synthase (85) and cellulose-like proteins (154). A combined approach of partial protein purification and proteomics was useful in the identification of a large family associated with the pectin biosynthetic HG:galacturonosyltransferases (137). The classification of GTs in the CAZY database was instrumental in identifying GT gene candidates targeted for a reverse genetic approach using the SALK T-DNA or other mutant library collections. Lastly, microarray analyses of tissues or cell types at different devel­opmental stages were useful in identifying secondary wall synthesis candidates (139, 509). Similar biochemical, genetic, and genomic approaches were successful in identifying biosyn­thetic genes involved in nucleotide-sugar synthesis and NDP-sugar transporters. The last decade has been a very fruitful and exciting time for the community of wall researchers (finally a crack in the wall).

Can bioinformatics assist in the prediction of GTs? Would the knowledge of UDP-binding sites in several plant GTs be useful to distinguish between GDP-, or ADP-transferases? Would the knowledge of the sugar moiety-binding site be similarly helpful to identify, for example, putative GalTs? Is the binding pocket for UDP-Rha in a flavonoid: RhaT the same as for pectin:RhaT? Clearly, a critical mass of biochemical knowledge is required to start to predict gene function by computer. We are not there yet__________________________

Acknowledgments

A review of this magnitude is not possible without the input, efforts, and patience of countless individuals. The authors express sincere gratitude to the multiple students, researchers, colleagues, and staff who contributed to this chapter through engaging discussions and editorial effort. Special thanks go out to our families who graciously endured our too frequent absences. We also thank the Department of Energy, the National Science Foundation and

the NRI, CSREES, USDA who provide the invaluable funding that supports plant cell wall research.