Detecting Nearly Periodically Expressed Genes from their Microarray Time-Course Profiles

L.-P. Tian (PR China), L.-Z. Liu, and F.-X. Wu (Canada)


time-course gene expression profiles, nearly periodically expressed gene, parameter estimation, F-testing


Periodic (more accurately, nearly periodic) biological processes play important roles in living organisms and can be studied at molecular level. Time-course gene expression profiles associated with those periodic biological processes appears nearly periodic. Detecting nearly periodically expressed genes from their microarray time-course data could help understand the molecular mechanism of those biological processes. This paper proposes a novel method for detecting periodically expressed genes from their time-course expression profiles. In the proposed method, a nearly periodical gene expression profile is modeled by a nearly periodic function, which is a linear combination of trigonometric sine function, cosine function, and a linear function in time variable plus a Gaussian noise term. As the model is nonlinear in parameters and time variable, a two-step parameter estimation method is employed for estimating parameters in the model. On the other hand, non periodical gene expression profiles are modelled by a constant plus a Gaussian noise term. The statistical F testing is used to make a decision if a gene is nearly periodically expressed or not. One synthetic dataset and two biological datasets were employed to evaluate the performance of the proposed method. The results show that the proposed method can effectively detect nearly periodically expressed genes.

Important Links:

Go Back