Logo Logo
Hilfe
Kontakt
Switch language to English
Methodological contributions to the challenges and opportunities of high dimensional clustering in the context of single-cell data
Methodological contributions to the challenges and opportunities of high dimensional clustering in the context of single-cell data
With the sequencing of single cells it is possible to measure gene expression of each single-cell in contrast to bulk sequencing which enables only average gene expression. This procedure provides access to read counts for each single cell and allows the development of methods such that single cells are automatically allocated to cell types. The determination of cell types is decisive for the analysis of diseases and to understand human health based on the genetic profile of single cells. It is of common use that cell types are allocated using clustering procedures that have been developed explicitly for single-cell data. For that purpose the single-cell consensus clustering (SC3), proposed by Kiselev et al. (Nat Methods 14(5):483-486, 2017) is part of the leading clustering methods in this context and is also of relevance for the following contributions. This PhD thesis aims at the development of appropriate analysis techniques for the clustering of high-dimensional single-cell data and their reliable validation. It also provides a simulation framework for the investigation of the influence of distorted measurements of single cells towards clustering performance. We further incorporate cluster indices as informative weights into the regularized regression, which allows a soft filtering of variables.
Non-linear embedding, Clustering, Internal validation, Simulation data, Single-cell RNA sequencing data
Fütterer, Cornelia Sigrid
2022
Englisch
Universitätsbibliothek der Ludwig-Maximilians-Universität München
Fütterer, Cornelia Sigrid (2022): Methodological contributions to the challenges and opportunities of high dimensional clustering in the context of single-cell data. Dissertation, LMU München: Fakultät für Mathematik, Informatik und Statistik
[thumbnail of Fuetterer_Cornelia.pdf] PDF
Fuetterer_Cornelia.pdf

3MB

Abstract

With the sequencing of single cells it is possible to measure gene expression of each single-cell in contrast to bulk sequencing which enables only average gene expression. This procedure provides access to read counts for each single cell and allows the development of methods such that single cells are automatically allocated to cell types. The determination of cell types is decisive for the analysis of diseases and to understand human health based on the genetic profile of single cells. It is of common use that cell types are allocated using clustering procedures that have been developed explicitly for single-cell data. For that purpose the single-cell consensus clustering (SC3), proposed by Kiselev et al. (Nat Methods 14(5):483-486, 2017) is part of the leading clustering methods in this context and is also of relevance for the following contributions. This PhD thesis aims at the development of appropriate analysis techniques for the clustering of high-dimensional single-cell data and their reliable validation. It also provides a simulation framework for the investigation of the influence of distorted measurements of single cells towards clustering performance. We further incorporate cluster indices as informative weights into the regularized regression, which allows a soft filtering of variables.