False Discovery Rate of Classification as a Function of Periodicity Strength of Time-Course Gene Expression
Farzad Najafi, Mahnaz Khalafi, Masoud Golalipour, Majid Azimmohseni
Department of Statistics, Faculty of Science, Golestan University, Gorgan, Iran
Medical cellular and molecular Research center, Golestan University of Medical Sciences, Gorgan, Iran
Abstract. Classification of genes provides valuable information about similar types of gene expressions. The periodic structure of time-course gene expression is a reliable characterization to classify two genes with the same periodic pattern in the same class. The strength of periodicity may differ from one gene to another. In this article, using Lomb-Scargle and JTK methods, three types of cyclic time-course patterns of genes are introduced according to periodicity strength. We proposed that the periodicity is an important factor for gene discrimination according to time-course expression profile. Based on the Saccharomyces cerevisiae data set, genes with different periodicity were discriminated, according to both the amounts of phase shift and time-course expression. Then, false discovery rates were computed under all circumstances. As a result, the false discovery rate increased when the strength of periodicity decreased. The false discovery rate of genes with strong periodic structure was 20% whereas it was 45% for weak periodic ones. The data set comprised 79% of genes with a weak periodicity, that deviated the result of discrimination.
Key words: discrimination, JTK, Lomb-Scargle method, phase shift, Saccharomyces cerevisiae data.