Abstract
With recent advances of microarray technology, we are able to monitor the expression of thousands of genes simultaneously. Detecting periodic signals from time-series microarray data is commonly used to facilitate the understanding of the critical roles and underlying mechanisms of regulatory transcriptomes. However, time-series microarray data is noisy. In particular, how the temporal data structure affects the performance of periodicity detection has remained elusive. In this dissertation, we address key issues in time-series gene expression analysis. We present novel computational methods based on empirical mode decomposition (EMD) to meet the challenges in three areas (1) a metric to measure the complexity of oscillatory nature of time-series, (2) an algorithm to search periodic patterns from time-series, and (3) protein complex co-expression networks of periodically expressed genes. We applied these methods to the yeast metabolic cycle (YMC) dataset to extract a series of intrinsic mode function (IMF) oscillations from the time-series data and to evaluate performance of periodicity detection methods. Our results show that 1,469 periodically expressed genes might have been under-detected in the original analysis because of interference between decomposed IMF oscillations. Most of the under-detected genes were mainly associated with ribosome biogenesis and RNA processing. Additionally, validated by our protein complex coexpression analysis, we confirmed that 56 genes were newly determined as periodic. We demonstrated that EMD can be used incorporating with existing periodicity detection methods to improve their performance. Our EMD approach can be applied to other time-series microarray studies. In addition, a number of recent studies have shown that loop-design is more efficient than reference control design. However, limited loop-design web-based tools are available. We have developed the THEME that exploits all necessary data analysis tools for loop-design microarray studies. This web platform provides data assessment and visualization tools to evaluate the performance of microarray experimental procedures. Data analysis procedures, starting from uploading raw data files to retrieving DEG lists, can be flexibly operated with natural workflows with THEME.