Flexible temporal expression profile modelling using the Gaussian process

Motivation: Time course gene expression experiments have proved valuable in a variety of biological studies (e.g., Chuang et al., 2002; Edwards et al., 2003). A general goal common to many of these time course experiments is to identify genes that exhibit different temporal expression profiles across multiple biological conditions. Such experiments are, however, often hampered by the lack of data analytical tools. It is our goal of the current paper to develop a rigorous yet flexible statistical method applicable in a wide range of time course experiments. Results: Taking advantage of the great flexibility of Gaussian processes, we propose a statistical framework for modelling time course gene expression data. It can be applied to both long and short time series and also allows for multiple differential expression patterns. The method can identify a gene’s temporal differential expression pattern as well as estimate the expression trajectory. The utility of the method is illustrated on both simulations and an experiment concerning the relationship between longevity and the ability to resist oxidative stress. Supplementary Information: The R-package for the proposed approach will be released at http://www.isye.gatech.edu/ ̃myuan/YuanBio.html Contact: [email protected]