A Workflow for the Analysis of DNA Microarray Time-Course Data

Abstract

The past two decades have witnessed the increasing use of high-throughputmeasurement technologies in biology and the advent of the –omics fields, including genomics,transcriptomics, proteomics, and metabolomics. These new measurement platforms havemotivated the development of novel data-analysis methods and workflows. Nowhere is thismore true than in transcriptomics, where DNA microarrays are widely used to measure geneexpression. One area that has suffered from a lack of development of new analysis tools is theapplication of DNA microarrays to time-course data. The use of DNA microarrays to followtemporal changes in biological systems is particularly important, allowing the measurement ofdynamic changes in gene expression and providing valuable insight into cellular regulation.However, there are many challenges to analyzing this type of DNA microarray data that aredistinct from other gene-expression experiments, thus necessitating the development of novelanalysis methods.This thesis reports the development of a workflow for the analysis of DNA microarraytime-course data. Particular emphasis is focused on the estimation and incorporation ofmeasurement uncertainties at each step, methods for data visualization and normalization, andthe decomposition of data using biologically meaningful models. The emphasis on measurementuncertainties led to a study of operator effects (gridding, flagging) on expression ratios, as wellas the validation of a bootstrap method to estimate measurement uncertainties in microarraydata. The application of correlation heat maps to time-course array level data allowed thevisualization and interpretation of transcriptome-wide changes in gene expression, providingpreliminary insights into the data. Microarray normalization was also investigated in the contextof time-course experiments, with a comparison of traditional and novel data normalizationmethods. Finally, the application and analysis of multivariate curve resolution using weightedalternating least squares (MCR-WALS) to time-course data is considered, with the extraction ofbiological information using the Gene Ontology. The biological systems investigated in this workinclude S. cerevisiae (yeast; cell cycle and exit from stationary phase), P. falciparum (malariaparasite; intraerythrocytic developmental cycle) and D. melanogaster (fruit fly; life cycle).Through the implementation of the workflow described in this thesis, putative regulatoryprofiles were extracted for each of these systems that were ontologically consistent with theknown biology.