Tyrosine sulfation is a ubiquitous PTM that predominantly modifies proteins in the secretory pathway (Bundgaard et al., 1995, 1997; Nicholas et al., 1999; Walsh and Jefferis, 2006), and plays an important role in regulating chemotaxis (Monigatti et al., 2006; Stone et al., 2009; Walsh and Jefferis, 2006), inflammatory response (Gao et al., 2003), and cell adhesion (Stone et al., 2009). In animals, the sulfation is catalyzed by two closely related tyrosylprotein sulfotransferases (TPST-1 and TPST-2) (Monigatti et al., 2006; Walsh and Jefferis, 2006), while a non-homologous AtTPST through convergent evolution has been identified in plants (Komori et al., 2009). In contrast with labor-intensive and time-consuming experimental assays, computational prediction of sulfation sites in proteins has become an efficient approach to generate useful information for further experimental verification. Previous studies suggested the short linear motif around the sulfation site is informative, and raised several consensus determinants for the prediction (Bundgaard et al., 1995, 1997; Monigatti et al., 2006; Nicholas et al., 1999). In 2002, Monigatti et al. presented the first online predictor of Sulfinator with four distinct Hidden Markov Models (HMMs) (Monigatti et al., 2002). With a Support Vector Machines (SVMs) classifier, Chang et al. developed SulfoSite for the prediction of sulfation sites (Chang et al., 2009). Recently, the algorithms of random forest (Yang, 2009) and nearest neighbor (Niu et al., 2010) were also adopted for predicting sulfation respectively, although the applicable tools were not released.

      In this work, we manually collected 273 experimentally indentified protein tyrosine sulfation sites in 171 unique proteins from scientific literature. A previously self-developed GPS (Group-based Prediction System) algorithm was employed with great improvement. We calculated the leave-one-out validation and 4-, 6-, 8-, 10-fold cross-validations to evaluate the prediction performance and system robustness. The leave-one-out validation result is accuracy (Ac) of 90.23%, sensitivity (Sn) of 89.60%, and specificity (Sp) of 90.36%. The online service and stand-alone packages of GPS-TSP 1.0 were implemented in JAVA 1.4.2 and freely available at: http://tsp.biocuckoo.org/.


GPS-TSP 1.0 User Interface

For publication of results please cite the following article:

Systematic analysis of the in situ crosstalk of tyrosine modifications reveals no additional natural selection on multiply modified residues. Zhicheng Pan, Zexian Liu, Han Cheng, Yongbo Wang, Tianshun Gao, Shahid Ullah, Jian Ren , Yu Xue.
Scientific Reports. 2014, 4: 7331
[Abstract] [FREE Full Text] [Free PDF] [Supplementary Data]