蛋白質複合體之研究-改善預測精確度與拓樸結構強韌性分析

Protein Complexes Study - Prediction Accuracy Improvement and Topological Robustness Analysis

周思瑜(Syz-Yu Chou)

指導教授 ： 黃建宏

共同指導教授 ： 吳家樂(Ka-Lok Ng)

國立虎尾科技大學/電機資訊學院/光電與材料科技研究所/碩士(2012年)

摘要

蛋白質複合體參與許多的生物過程，在生物體中扮演不可或缺的角色。目前預測蛋白質複合體的方法大都是植基於假設蛋白質交互作用的密集區構成蛋白質複合體，本論文中我們探討加入物理化學性質的考量以改進蛋白質複合體預測的準確性。
我們使用主成分分析評估主要特徵，並且使用十倍交叉驗證來評估四種機器學習方法－支持向量機、類神經網路、決策樹、貝式分類器的分類準確度。
我們進一步評估使用物理化學特性在預測蛋白質複合體效能，我們發現使用胺基酸組成作為主要特徵，結合支持向量機所建置的預測機制，可以有效的改進預測準確度。我們也將研究結果建置成一個網站提供使用者預測蛋白質複合體，網址為 http://bioinfo.csie.nfu.edu.tw:8080/ProteinComplex/ProteinComplexsvm.aspx。
在系統生物學中生物網路的強韌性探討也是重要的議題。本論文中我們延伸蛋白質交互作用網路考慮蛋白質複合體之間的蛋白質交互作用，建構蛋白質複合體網路，結果顯示除了移除最大分支度的點的擾動攻擊外，人類與酵母菌蛋白質複合體網路相對於其他三種擾動攻擊方法都相當穩定。

關鍵字

蛋白質複合體 ； 蛋白質交互作用 ； 機器學習 ； 胺基酸組成 ； 物理化學屬性 ； 網路強韌性

並列摘要

It is known that protein complexes are involved in many biological processes. Most of the protein complex prediction algorithms are based on the assumption that protein-protein interaction (PPI) dense regions can possibly lead to complexes formation. In this study we propose the inclusion of physiochemical properties is necessary for improving protein complex prediction. Principle component analysis is carried out to determine the major features. Cross-validation test is performed to test the classification accuracy of four machine learning methods; i.e. support vector machining, neural network, decision tree and Bayes classifier.
To study the effectiveness of adopting physiochemical properties on protein complex prediction, we shown that prediction accuracy can be improved post-processing prediction methods’ results with an amino acid composition profile. A web service for protein complex prediction has been set up, which can be accessed at http://bioinfo.csie.nfu.edu.tw:8080/ProteinComplex/ProteinComplexsvm.aspx
The question of the robustness of a biological network upon perturbation is an important issue in systems biology study. In this thesis, we extend the idea of protein-protein interaction network by considering interactions among protein complexes, and construct the so-called protein complex networks (PCNs). Stability of two species’ PCNs is studied. Except the attack perturbation, the PCNs of yeast and human species are quite robust with respect to failure, rewiring and edgedeletion perturbations.

並列關鍵字

protein complex ； protein-protein interaction ； Gene Ontology ； sequence alignment ； physiochemical property ； amino acid composition profile ； machine learning method ； netword robustness

參考文獻

[1] W. J. Henzel, T. M. Billeci, J. T. Stults, S. C. Wong, C. Grimley, and C. Watanabe. (1993). Identifying Proteins from Two-Dimensional Gels by Molecular Mass Searching of Peptide Fragments in Protein Sequence Databases. Proc. Natl. Acad. Sci. USA 90, pp. 5011-5015.

[4]X. Li, M. Wu, C. K. Kwoh and S. K. Ng. (2010). Computational Approaches for Detecting Protein Complexes from Protein Interaction Networks: A Survey. BMC Genom, 11(suppl1):S3, pp. 1-19.

[5]M. Wu, X. Li, C. K. Kwoh and S. K. Ng. (2009). A Core-Attachment Based Method to Detect Protein Complexes in PPI Networks. BMC Bioinfo., 10:169, pp. 1-16.

[9]C. H. Huang, J. F. Fang, J. P. Tsai, and K. L. Ng. (2006). Topological Robustness of the Protein-protein Interaction Networks, Lecture Notes in Bioinformatics, vol. 4023, pp. 166-177.

[10]N. Barkai, and S. Leibler. (1997). Robustness in simple biochemical networks, Nature, vol. 387, 1997, pp. 913-917.