Sten Ilmjärv „Estimating differential expression from multiple indicators“
Thesis supervisor: prof Jaak Vilo, PhD, Institute of Computer Science; prof Eero Vasar, PhD, Institute of Biomedicine and Translational Medicine; Senior Research Fellow Hendrik Luuk, PhD, Institute of Biomedicine and Translational Medicine.
Oponent: Senior Research Fellow Djork-Arné Clevert, PhD, Institute of Bioinformatics, Johannes Kepler University, Linz, Austria.
Summary
With very few exceptions the genome is identical in every cell of our body. Nevertheless, our body consists of many different cells with unique shape, physiology and behavior. This variation in cell types is achieved by coordinated activity of every gene in the genome. In other words, every cell has it’s own program that regulates the activity of the genes, which subsequently determines the cell’s morphology and functionality.
There are several technologies that can be used for measuring the activity of a gene i.e. gene expression. One of the best-known and widely used technologies is DNA-microarray developed more than two decades ago. The advantage of microarrays is that they can measure gene expression of tens of thousands of genes simultaneously. Additionally, good standards for data housing and data analysis have been established and currently the amount of microarray data in public warehouses from experiments measuring gene expression is unmatched. Therefore the reuse of available data to find answers to alternative hypothesis and the development of added-value bioinformatics tools requires flexible solutions that best exploit the potential hidden in the data.
This work focused on developing a new framework for differential expression analysis of microarray data, termed Differential Expression from Multiple Indicators (DEMI). Our motivation was to utilize the information stored in the repeated measurements on microarrays (a gene’s expression is measured from multiple locations) to increase the sensitivity of the analysis. In comparison to other well-established methods, DEMI demonstrated a good and stable performance regardless of the microarray platform and sample size. This is especially important when samples are hard to obtain, like in clinical trials, or due to limitation in the resource, which is often the case in pilot studies. Furthermore, DEMI can handle experiments with non-conventional design, like time- or dose-dependent differential expression analysis with no replicates and identify genomic regions with unidirectional changes in gene expression levels of neighboring genes (e.g. a decrease in gene expression levels of neighboring genes can result due to large-scale epigenetic changes caused by a cancer). All in all, DEMI provides a flexible solution for differential expression analysis and is able to answer new questions from already published data.