Tyrosine sulfation is a post-translational modification occurring on many secreted and membrane-bound proteins. Previous studies have implicated tyrosine sulfation as a determinant of protein-protein interactions involved in leukocyte adhesion, hemostasis and chemokine signalling. Due to the biological significance, several studies have been proposed to analyze the consensus sequences of tyrosine sulfation sites. Nevertheless, they did not take the consideration of protein structural information. Therefore, the intrinsic feature of sulfated protein is still elusive and remains to be delineated. In this work, we present SulfoSite, a novel computational method based on support vector machine (SVM) for prediction of sulfotyrosine sites. The method was developed to consider the structural information such as solvent accessibility surrounding the tyrosine sulfation sites. 162 experimentally verified tyrosine sulfation sites were extracted from Swiss-Prot release 53.0, then removed the homologous sequences with a given window size. Following evaluation of 5-fold cross-validation, the results indicates that the value of solvent accessibility surrounding the sulfotyrosine sites can improve the prediction accuracy. The SVM classifier was able to accomplish an accuracy of 96 %, a sensitivity of 81 %, and a specificity of 99 % in the independent test, respectively. This method significantly outperforms previous attempts of tyrosine sulfation sites predictions. Taken together, our analyses demonstrate that SulfoSite is a useful computational program for sulfotyrosine prediction. The effective SulfoSite is freely accessible from http://SulfoSite.mbc.nctu.edu.tw.