In the post-genomic era, elucidation of different types of post-translational modifications (PTMs) of proteins is fundamental for understanding the dynamics of proteome and various cell signaling pathways/networks. As one of the most important and universal PTMs, protein ubiquitination is a reversibly biochemical process, covalently forming an isopeptide bond between the C-terminal –GG carboxy group of a ubiquitin protein and the ε-amino group of lysine residues or, less commonly, other types of residues of a substrate protein (Aguilar and Wendland, 2003; Haglund, et al., 2003; Hicke, 2001; Hicke and Dunn, 2003; Pickart, 2001; Reinstein and Ciechanover, 2006; Semple, 2003; Sun and Chen, 2004; Wong, et al., 2003). Ubiquitination regulates a variety of cellular processes, including cell division/mitosis, signal transduction, endocytosis, and apoptosis, etc (Aguilar and Wendland, 2003; Haglund, et al., 2003; Hicke, 2001; Pickart, 2001; Sun and Chen, 2004). And aberrant of the ubiquitin-proteasome system (UPS) is implicated in numerous pathologies, such as neurodegenerative disorders, inflammatory diseases and cancers, etc (Hoeller, et al., 2006; Reinstein and Ciechanover, 2006; Wong, et al., 2003).
    In this field, identification of Ubiquitinated proteins with their sites is one of the greatest challenges and important for understanding the molecular mechanism of ubiquitin system and regulatory roles of ubiquitination. Besides the conventional experimental approaches, such as site-mutagenesis of potential ubiquitination sites (Lin, et al., 2005), antibodies of Ub (anti-Ub) (Gentry, et al., 2005), high-throughput methods with mass-spectrometry (MS) have also been employed (Kirkpatrick, et al., 2005). Currently, the large-scale identification of Ubiquitinated proteins is mainly focused on Saccharomyces cerevisiae (Hitchcock, et al., 2003; Mayor, et al., 2005; Peng, et al., 2003). More than 1,000 proteins have been identified as potential ubiquitinated targets (Hitchcock, et al., 2003; Mayor, et al., 2005; Peng, et al., 2003). However, only ~140 ubiquitination sites were precisely mapped (Hitchcock, et al., 2003; Peng, et al., 2003). Although identification of ubiquitination sites could be performed experimentally, it's time-consuming and labor-intensive. Therefore, computational prediction of ubiquitination sites is desirable for its convenience and fast-speed.
    Compared to other several common PTMs, e.g., phosphorylation (Xue, et al., 2006; Xue, et al., 2005), palmitoylation (Xue, et al., 2006; Zhou, et al., 2006) and sumoylation (Xue, et al., 2006; Zhou, et al., 2005), computational studies of ubiquitination sites attract less attention. To our knowledge, there is only one article to compute the preference of ubiquitination sites with their flanking short peptides in yeast (Catic, et al., 2004). The results provided clear sequence and 3D structural preferences for ubiquitination. Thus, the ubiquitination sites is theoretically predictable, although the data set was limited (135 sites) at that time. In this work, we present a novel online web server for protein ubiquitination site prediction called BDM-PUB, Prediction of Ubiquitination site with Bayesian Discriminant Method. We have manually mined scientific literature to collect 279 experimentally verified ubiquitination sites of 162 distinct proteins. After redundant-clearing, there are 260 sites from 154 substrates reserved. Then the BDM (Bayesian Discriminant Method) algorithm (Xue, et al., 2006) has been employed. The prediction performance of BDM-PUB is satisfying with high accuracy.