The past decade has witnessed the rapid progresses on functional dissections of protein sumoylation (Geiss-Friedlander and Melchior, 2007). The SUMO (small ubiquitin-related modifier) gene SMT3 was firstly identified in S. cerevisiae as a suppressor of the centromeric protein Mif2 (Meluh and Koshland, 1995), and later was shown to be covalently coupled to the Ran GTPase-activating protein RanGAP1 as a reversible modifier (Mahajan, et al., 1997; Matunis, et al., 1996). Proteins modified by SUMO could alter their sub-cellular localization, activity or stability, etc (Fernandez-Lloris, et al., 2006; Mahajan, et al., 1997; Matunis, et al., 1996). And protein sumoylation plays important roles in a variety of biological processes, such as transcriptional regulation, signaling transduction, cell cycle progression and differentiation (Deyrieux, et al., 2007; Gill, 2004; Montpetit, et al., 2006; Seeler and Dejean, 2003), etc. In addition, aberrance of SUMO system is highly implicated in numerous diseases and cancer developments (Dorval and Fraser, 2007; Fernandez-Lloris, et al., 2006; Li, et al., 2005; Seeler, et al., 2007).

    In this work, we updated our SUMOsp 1.0 into version 2.0. The training data set was manually collected from scientific literature. The non-redundant training data contained 279 sumoylation sites from 166 distinct proteins. Then an updated version of GPS algorithm was deployed. The self-consistency, leave-one-out validation and 4-, 6-, 8-, 10-fold cross-validations were calculated to evaluate the prediction performance and system robustness of SUMOsp 2.0. Also, the prediction performance was tested on an additional data set not included in the training data set, with 53 sumoylation sites from 31 proteins. We compared SUMOsp 2.0 with SUMOplot and SUMOsp 1.0, on both the training data and new data. The specificity (Sp) of SUMOsp 2.0 was improved significantly, while the sensitivity (Sn) was similar or just slightly reduced against previous tools. The SUMOsp 2.0 was implemented in JAVA 1.4.2 and would use local CPU for computation. With a high speed, SUMOsp 2.0 could predict out potential sumoylation sites for ~1,000 proteins (with an average length of ~1000aa) within ten minutes. Taken together, we proposed that the highly specific SUMOsp 2.0 web server will be more efficient for sumoylation sites prediction. The SUMOsp 2.0 is freely available at: http://bioinformatics.lcd-ustc.org/sumosp.

    This website is linked in ExPASy Proteomics Tools page.

SUMOsp 2.0 User Interface