Identification of Secretory Proteins by separated space based linear discriminant analysis
Signal peptides are short regions of amino acid residues, which have become a crucial tool in finding new drugs or reprogramming cells for gene therapy. Owing to the rapidly increasing number of protein, it is highly demanded to develop the automated algorithm to identify the signal peptides. Recently, we had adopted a new alignment kernel function to identify secretory proteins. Compared with previous works, our method improves the predictive performance and is much more stably. However, we also find it will be more helpful to visualize the classification. Study on feature reduction and extracting the useful features for classification, we make full use of the null space of within-class scatter matrix, and propose separated space based linear discriminant analysis(SSLDA). For signal peptides, with the high-dimension got by indefinite kernel based on global alignment similarity, we apply SSLDA and get reduced feature, then sequential data can be visualized, which are highly demanded in machine learning, and avoid the lack of physical explanation as classical neural network method did. The classification results also prove the performance of SSLDA.
signal peptides predict separated space based linear discriminant analysis visualize
Hui Liu Yuan Yao Xiang Liu Kuo-Chen Chou
Department of Biomedical Engineer Dalian University of Technology Dalian, China The second department PLA Commanding Communications Academy Wuhan, China School of Materials Science and Engineering Dalian Jiaotong University Dalian, China Gordon Life Science Institute San Diego, CA 92130, USA
国际会议
上海
英文
1979-1983
2008-05-16（万方平台首次上网日期，不代表论文的发表时间）