Identifying Periodically Expressed Genes Based On Gene Expression Profile Data

The development of DNA microarray technology makes it possible to study the mechanism of gene expression on a genome-wide scale. The technology is used in cell cycle experiments to get gene expression data sets for gene expression analysis. The studies find that many genes show significantly periodical expression during the cell cycles, and these genes play important roles in the cell cycle regulation. According to the analysis of periodical genes, the mechanism of gene regulation can be realized. As a result, how to identify the periodically expressed genes becomes a challenge.Recently, many methods have been developed to identify periodically expressed genes based on expression profile data sets. Almost all the methods use a model based on some kinds of periodical functions, so these methods can not describe the expression patterns of genes essentially. This paper does not use a periodical function, but compares the similarity of time-course expression data sets in two continuous cell cycles. The similarity is measured by the correlation coefficient in linear regression analysis, and it is defined as the major feature to identify periodically expressed genes. Considering the misjudgment of correlation coefficient, this paper introduces the fluctuation feature measured by variance to help identifying periodically expressed genes.Three time-course expression data sets are used to identify the periodically expressed genes. As a result, 989 genes are identified to be potential periodically expressed genes, including 95 genes which are known to be periodically expressed genes, and the accuracy reaches 84%.