Exploiting maximal dependence decomposition to identify S-sulfenylation sites with substrate motifs

• Main Authors: van-minh bui, 裴文明
• Other Authors: Tzong-Yi Lee
• Format: Others
• Language: en_US
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/22na83

碩士 === 元智大學 === 資訊工程學系 === 104 === S-sulfenylated protein, the covalent attachment of a hydroxyl group to a thiol of a cysteine amino acid, is a type of post-translational modification (PTM). This protein plays a crucial role in various biological processes including transcriptional regulation, apoptosis and cytokine signaling. However, the shortage of discoveries in S-sulfenylation protein has many limited for researchers. The issues in discrimination the substrate site of this protein consider being essential tasks in computation biology for further studies in protein structures and functions. Most of methods are used to identify this protein to be done in practical laboratories, so the cost is quite expensive to buy complex equipment and sample and the execute time takes so much time for many steps. Until now, there are no tools dedicated to the computational identification of SOH sites. With a total of 1096 experimentally S-sulfenylated proteins, the study was carried out a bioinformatics investigation on SOH sites based on amino acid composition. A TwoSampleLogo indicates that the positively and negatively charged amino acids flanking the SOH sites may impact the formulation of S-sulfenylation. Moreover, the study also invested the potential features to build the best model in order to identify effectively the cysteine residue of this protein. In addition, the maximal dependence decomposition (MDD) was utilized to explore the substrate motifs of SOH sites. Based on the concept of binary classification between SOH and non-SOH sites, Support Vector Machine (SVM) was applied to learn the predictive model from MDD-identified substrate motifs. According to the evaluation of five-fold cross validation, the integrated SVM model learned from substrate motifs yields an average accuracy of 0.87, that leads to a significant improvement in the prediction of SOH sites. Moreover, the integrated SVM model also efficaciously enhances the performance in an independent testing set. Finally, the integrated SVM model was applied to implement an effective web resource, named MDD-SOH, for identifying SOH sites with their corresponding substrate motifs (http://csb.cse.yzu.edu.tw/MDDSOH/index.php).