带噪生物实验数据的不确定支持向量数据描述  

Uncertainty Support Vector Domain Description for Biological Experiment Data with Noise

作　　者：卜令超[1] 王士同[1] 

机构地区：[1]江南大学信息工程学院,江苏无锡214036

出　　处：《生物信息学》2009年第4期295-299,共5页Chinese Journal of Bioinformatics

摘　　要：在对候选基因进行排序时,支持向量数据描述(SVDD)可以用来描述各种异构的数据源,如序列数据、学术文献数据、各种生物实验数据等。由于生物实验数据带有噪声,在用SVDD对其描述时,会遇到噪声的影响。本研究通过公式推导扩展了原始的SVDD,提出不确定支持向量数据描述(USVDD),用来降低噪声的影响。利用酵母基因表达数据进行实验,结果表明该方法比标准的SVDD对带噪声的数据具有更好的描述能力。Support vector domain description （SVDD） method can be used to describe each heterogeneous data source for the prioritization of candidate genes, for example, the sequence information, the published literature, kinds of biological experiment data and so on. However, when SVDD is used for the biological experiment data, which is usually corrupted with noise, the result will be greatly affected by the noise. Uncertainty support vector domain description （USVDD） is presented for de- noising, through new formulations expanding the original SVDD. The results of experiments with yeast gene expression data indicated that the proposed method here is superior to the standard SVDD for noisy data.

关 键 词：支持向量数据描述 不确定性 基因表达数据 降噪 支持向量机 

分 类 号：Q811[生物学—生物工程]