In proteomics, a proteinpsilas function is always strongly related to its structure. But, while some parts of a protein have a fixed definite structure, such as alpha-helix, beta-sheet, or coil, other parts are not associated with well-defined conformations. Previously, these so-called disordered regions were not thought to have a specific function of their own. But, recent studies suggest that some disordered regions may have important signaling or regulatory functions. In addition, some critical diseases are strongly related to these disordered regions. Hence, prediction of these disordered regions is essential. In this paper, we try to use the support vector machine (SVM) to predict the disordered regions. Furthermore, this paper emphasizes post processing of the SVM prediction results. Two post-processing algorithms are introduced. These algorithms are used to smooth the primary results by SVM. Different from other studies, these smoothing steps are related to the neighborspsila distance to the candidate node. The results show that these algorithms can improve the prediction accuracy further by 1%.
關聯:
Proceedings of the 2nd International Conference on Complex, Intelligent and Software Intensive Systems (CISIS 2008), pp.693-698