Predicting sumoylation sites using support vector machines based on various sequence features, conformational flexibility and disorder

Warning The system is temporarily closed to updates for reporting purpose.

Yavuz, Ahmet Sinan and Sezerman, Uğur (2014) Predicting sumoylation sites using support vector machines based on various sequence features, conformational flexibility and disorder. BMC Genomics, 15 (Supplement: 9). ISSN 1471-2164 (Print) 1471-2164 (Online)

[img]PDF (Listed in DOAJ as an open access journal) - Requires a PDF viewer such as GSview, Xpdf or Adobe Acrobat Reader

Official URL: http://dx.doi.org/10.1186/1471-2164-15-S9-S18


Background Sumoylation, which is a reversible and dynamic post-translational modification, is one of the vital processes in a cell. Before a protein matures to perform its function, sumoylation may alter its localization, interactions, and possibly structural conformation. Abberations in protein sumoylation has been linked with a variety of disorders and developmental anomalies. Experimental approaches to identification of sumoylation sites may not be effective due to the dynamic nature of sumoylation, laborsome experiments and their cost. Therefore, computational approaches may guide experimental identification of sumoylation sites and provide insights for further understanding sumoylation mechanism. Results In this paper, the effectiveness of using various sequence properties in predicting sumoylation sites was investigated with statistical analyses and machine learning approach employing support vector machines. These sequence properties were derived from windows of size 7 including position-specific amino acid composition, hydrophobicity, estimated sub-window volumes, predicted disorder, and conformational flexibility. 5-fold cross-validation results on experimentally identified sumoylation sites revealed that our method successfully predicts sumoylation sites with a Matthew's correlation coefficient, sensitivity, specificity, and accuracy equal to 0.66, 73%, 98%, and 97%, respectively. Additionally, we have showed that our method compares favorably to the existing prediction methods and basic regular expressions scanner. Conclusions By using support vector machines, a new, robust method for sumoylation site prediction was introduced. Besides, the possible effects of predicted conformational flexibility and disorder on sumoylation site recognition were explored computationally for the first time to our knowledge as an additional parameter that could aid in sumoylation site prediction.

Item Type:Article
Additional Information:Article Number: S18
Uncontrolled Keywords:Sumoylation; SUMO; machine learning; support vector machines; post-translational modification
Subjects:Q Science > Q Science (General)
ID Code:26224
Deposited By:Uğur Sezerman
Deposited On:12 Dec 2014 21:22
Last Modified:02 Aug 2019 15:39

Repository Staff Only: item control page