Ding Hui, Liang Zhi-Yong, Guo Feng-Biao, Huang Jian, Chen Wei, Lin Hao
Key Laboratory for NeuroInformation of Ministry of Education, Center of Bioinformatics, School of Life Science and Technology and Center for Information in Biomedicine, University of Electronic Science and Technology of China, Chengdu 610054, China.
Key Laboratory for NeuroInformation of Ministry of Education, Center of Bioinformatics, School of Life Science and Technology and Center for Information in Biomedicine, University of Electronic Science and Technology of China, Chengdu 610054, China.
Comput Biol Med. 2016 Apr 1;71:156-61. doi: 10.1016/j.compbiomed.2016.02.012. Epub 2016 Feb 26.
A bacteriophage is a virus that can infect a bacterium. The fate of an infected bacterium is determined by the bacteriophage proteins located in the host cell. Thus, reliably identifying bacteriophage proteins located in the host cell is extremely important to understand their functions and discover potential anti-bacterial drugs. Thus, in this paper, a computational method was developed to recognize bacteriophage proteins located in host cells based only on their amino acid sequences. The analysis of variance (ANOVA) combined with incremental feature selection (IFS) was proposed to optimize the feature set. Using a jackknife cross-validation, our method can discriminate between bacteriophage proteins located in a host cell and the bacteriophage proteins not located in a host cell with a maximum overall accuracy of 84.2%, and can further classify bacteriophage proteins located in host cell cytoplasm and in host cell membranes with a maximum overall accuracy of 92.4%. To enhance the value of the practical applications of the method, we built a web server called PHPred (〈http://lin.uestc.edu.cn/server/PHPred〉). We believe that the PHPred will become a powerful tool to study bacteriophage proteins located in host cells and to guide related drug discovery.
噬菌体是一种能够感染细菌的病毒。被感染细菌的命运由位于宿主细胞内的噬菌体蛋白质决定。因此，可靠地识别位于宿主细胞内的噬菌体蛋白质对于理解其功能以及发现潜在的抗菌药物极为重要。因此，在本文中，我们开发了一种仅基于氨基酸序列来识别位于宿主细胞内的噬菌体蛋白质的计算方法。我们提出将方差分析（ANOVA）与增量特征选择（IFS）相结合来优化特征集。使用留一法交叉验证，我们的方法能够区分位于宿主细胞内的噬菌体蛋白质和不位于宿主细胞内的噬菌体蛋白质，最大总体准确率为84.2%，并且能够进一步将位于宿主细胞质和宿主细胞膜内的噬菌体蛋白质进行分类，最大总体准确率为92.4%。为了提高该方法实际应用的价值，我们构建了一个名为PHPred的网络服务器（〈http://lin.uestc.edu.cn/server/PHPred〉）。我们相信PHPred将成为研究位于宿主细胞内的噬菌体蛋白质以及指导相关药物发现的有力工具。
