In recent years the possibility to engineer cells in vitro has encountered significant progress. This engineering of cellular states, which is tightly coupled to the field of stem cell research, is considered to be a very useful technology for the generation of specialised cells for drug development, disease modelling, or regenerative medicine. One important part of this process is the quality control, i.e. the detailed characterisation of the end products, to ensure that the transformed cells are similar to their in vivo counterparts. Many markers and functional assays exist that can be used for quality control of these cells. However, most of them focus on specific, relatively narrow properties of the cells, neglecting a global overall comparison to the desired cell type.Here, we present a genome-wide gene expression microarray based approach to cell characterisation, providing complementary information to the commonly used single gene or morphological markers. We use a dimension reduction approach to localise newly generated microarray data in the high dimensional expression space. Using a combination of unsupervised and supervised dimension reduction methods, we establish a two-scale map of global gene expression with phenotypic interpretation of the coordinates.This two-scale map is used to characterise several different samples. It is first validated on a dataset of 24 different tissues and cell lines as well as on two datasets of artificially mixed tissues. Using these datasets, it is shown that the developed method outperforms three existing methods for RNA based global cell characterisation and that it provides increased information compared to the purely unsupervised or purely supervised dimension reduction methods. Application of the two-scale map to characterise in vitro transformed cells prooves to be useful in providing complementary information to the typical marker based or morphological criteria. In this respect, we could identify two examples of in vitro transformed cells where the transformation process is incomplete on a global expression level. Furthermore, we can show that in vitro differentiation of pluripotent stem cells results in immature cells that are similar to embryonic of fetal tissues of the respective type.Using microarray data from artificial mixtures of different tissues, we can observe clear non-linear effects in the data that fit well to the current understanding of the relationship between the RNA content of cells and the measurement signal of microarrays. Such non-linear effects are currently not captured by the proposed linear dimension reduction approach and give important hints for further improvements of the method.In addition to quality control of in vitro transformed cells, the two-scale decomposition approach developed in this thesis may also be useful for a number of other applications, such as the analysis of drug response profiles or disease progression.
|Qualification||Doctor of Philosophy|
|Award date||13 Jul 2015|
|Publication status||Published - 2015|