Bioinformatic Analysis of Cancer using Single-Cell Expression Data

Cancer is a highly diverse set of diseases, which despite sharing key characteristics, such as the ability to escape immune destruction, also vary both between patients and tissues - and sometimes even within a single tumor. To provide optimal treatment, an increased understanding of cancer is necessary. This can be achieved through characterization of malignant cells as well as their interaction partners, such as the cells of the immune system.
This thesis is comprised of five different studies, which are focused on cancer analysis using bioinformatics. The first study is centered around currently applied protocols for clinical cancer subtyping, which include bulk expression profiling of the cancer tissue. A challenge with the existing subtyping tools is their instability, but the work from this thesis demonstrates several approaches to overcome this to enhance future patient stratification. The second study from this thesis elaborates on the general idea of clinical cancer analysis, by surveying if single-cell data provides useful insights for primary breast cancer cases. Specifically, this study identifies a problem in the use of existing subtyping approaches for single-cell RNA-sequencing data, which is hypothesized to be caused by data sparsity. Additionally, the work also highlights intra-tumor heterogeneity patterns with potential clinical relevance, although further studies are needed to establish the validity of these results.
The third study of this thesis presents the computational tool cyCombine, which enables integration of cytometry datasets by providing frameworks for both batch correction and panel merging. The paper introducing this tool further demonstrates its utility by analyzing immune cells in relation to chronic lymphocytic leukemia, where changes in the CD8+ T cell states in cancer patients relative to healthy donors are detected.
The fourth study presents a transcriptome-based analysis of two Merkel cell carcinoma tumors, which demonstrates a reduction of antigen presentation gene expression. Finally, the fifth study introduces analysis of the immune infiltrate of clear cell renal cell carcinoma tumors using mass cytometry, validating that advancing disease can be coupled to immune dysfunction.
In combination, this thesis presents bioinformatic approaches to leverage single-cell RNA and protein expression data in cancer research across both hematological and solid malignancies. Not only does this involve utilization of existing tools to derive biological conclusions, but a framework for subtyping as well as a tool for data integration are also presented to aid future studies in the field.