| Περίληψη: | The process of embryogenesis is governed by the expressions of groups of genes which are acting in a coordinate way. Uncovering how the expressions of these genes control the development of a multicellular organism is fundamental for developmental biology.
Gene expression atlases could capture quantitative spatio-temporal information of all genes expressed in a developing embryo. In other words, they could reveal the underlying genetic activity during the embryogenetic processes by providing information about, apart from how much and which, where and when the genes are expressed at cellular level and the interactions between them. The extraction of relative gene expression data and their simultaneous study in animal models, such as zebrafish, would be greatly facilitated by the existence of such atlases.
This Master Thesis focuses on the early zebrafish embryogenesis and its goal is to design and implement an image processing framework that will provide the means to gather the expression patterns of different genes from different embryos at a given developmental stage into a common template. The framework should work with image-based data from different embryos, each fluorescently stained to label nuclear DNA and the expression patterns of a reference gene and another gene of interest. The volumetric data from each fluorescent label are contained in different channels. Therefore the crux of the framework lies in its ability to combine appropriately the information from the different channels and deal with a three-dimensional image registration problem.
The implemented framework works with datasets of two embryos, one that serves as a template and depicts a whole embryo and another that has to be aligned with the first and depicts part of the other embryo. It is composed of different steps, responsible for preprocessing the channels, coarsely positioning the partial embryo view in the three dimensional template’s space and determining the geometrical transformation that finally aligns it with the template using as reference the gene expression pattern common to all labelled embryos. The resulting transformation is used to map the second expression patterns, thus producing their spatial expression atlas. The algorithm developed is based on the Insight Segmentation and Registration Toolkit.
The framework was evaluated with data from six embryos at the same developmental stage and four different registration methods were compared in terms of performance. Visual inspection of the results identified the combination of the correlation coefficient, as a similarity measure function between two images, with the gradient descent optimization algorithm as the most appropriate method for this specific application. The final results obtained showed that the framework achieves its goal of integrating several gene expression patterns into a common template.
This framework is ready to be used in order to construct a gene expression atlas integrating a large number of gene expression data of the zebrafish embryogenesis. In the near future, this atlas should be validated with known genetic interactions and used to unravel new ones, so that gene regulatory network models can become a reality.
|
|---|
