Gene Filtering for Time Course Gene Expression Data Using the Growth Curve Model

• Activity Number: 76
• Type: Contributed
• Date/Time: Sunday, August 4, 2013 : 4:00 PM to 5:50 PM
• Sponsor: Biometrics Section
• Author(s): Sayantee Jana*+ and Narayanaswamy Balakrishnan and Dietrich von Rosen and Jemila Hamid
• Companies: McMaster University and McMaster University and Swedish University of Agricultural Sciences and McMaster University
• Keywords: Generalized multivariate analysis of variance ; growth curve model ; high-dimensional data ; moderated trace test ; Moore-Penrose generalized inverse ; multivariate bias and mean square error

Abstract

In time course microarray experiments the usual method of gene-filtering using dispersion measures are not applicable. A method based on multivariate growth curve model, that takes the within correlation and temporal ordering into account, is proposed. Two types of moderations for the variance-covariance matrix are considered to address the issue of high-dimensionality: the Moore-Penrose generalized inverse and Stein's shrinkage estimator. Extensive simulations demonstrated that our moderated test is unbiased, and is monotone with respect to sample size and parameter value regardless of the covariance structure. We also illustrated our method using a lung cancer data, where measurements were taken from normal and chemically treated human lung tissues at several time points. From a pool of 22,277 genes, 1053 genes (4.7%) were filtered out as non-noise genes. This is in sync with most biological experiments, where around 5% genes are found to be expressed. Although the focus of this study was on hypothesis testing, moderated MLE for the model parameter matrix was also provided and its bias and MSE were investigated empirically.