Blasi, Thomas
[Author]
;
Theis, Fabian J. ( )
[Contributor];
Theis, Fabian J. ( );Gagneur, Julien ;Marioni, John (Prof., Ph.D.)
[Contributor]
Data-driven statistical learning to model cellular heterogeneity ; Daten-getriebene Modellierung von Heterogenität zwischen Zellen mittels Methoden aus der Statistik und dem maschinellen Lernen
You can manage bookmarks using lists, please log in to your user account for this.
Media type:
Electronic Thesis;
Doctoral Thesis;
E-Book
Title:
Data-driven statistical learning to model cellular heterogeneity ; Daten-getriebene Modellierung von Heterogenität zwischen Zellen mittels Methoden aus der Statistik und dem maschinellen Lernen
Contributor:
Blasi, Thomas
[Author]
imprint:
Technical University of Munich; Technische Universität München, 2016-09-29
Footnote:
Diese Datenquelle enthält auch Bestandsnachweise, die nicht zu einem Volltext führen.
Description:
In the last decades the advent of new experimental techniques has lead to a drastic increase of available data in biology. As a consequence the importance of mathematical methods to deduct scientifically relevant hypothesis from this big amount of data is steadily growing. A major challenge for bio-mathematics and bio-statistics therefore lies in both the adaption of existing methods to the, often very specific, properties of the measured data, and in the development of new methods to model these data. In this thesis we present methods from statistics and machine learning that are suitable to perform this task. The quest for new mathematical methods, thereby, is always pursued in conjunction with the goal to find new scientific insights into the investigated biological system. The biological focus of this work is the analysis of heterogeneity among cells: almost all cells of a living organism share the same DNA, yet there is a multitude of different cell types that may all perform different tasks within the organism. The aim of this thesis is to explore both the biological principles that lead to cellular heterogeneity, and to improve the identifiability of different cellular phenotypes with mathematical methods. For this purpose four different mathematical methods are implemented, tested and applied to biological data in order to draw new conclusions about cellular heterogeneity: (i) We propose a statistical method to correct for latent confounding effects on single cell transcriptomics data that are due to differences in cell size, which we show to have an impact on the inference of the underlying gene expression mechanism. (ii) By applying ordinary-differential-equation-based models on chromatin data we can show that histone acetylation (a certain class of chromatin modifications with known impact on transcriptional regulation) depends specifically on the chromatin status before these modifications occur. (iii) We apply transfer entropy to protein time-series data from hematopoietic stem and progenitor cells ...