| Summary: | <p>Cells are often going through many complex dynamical changes, which change their function and transcriptional profile. It is often the case that this process is supported or initiated by neighboring cells in the tissue. The goal of this thesis is to better understand the neighborhood composition of cells and quantitatively describe its effects on the cells' transcriptional activity.</p>
<p>Single-cell RNA sequencing methods have proven valuable in providing high resolution, high throughput and unbiased whole-transcriptome gene expression measurements in individual cells. However, the information about the spatial proximity between cells is lost in single-cell sequencing experiments due to tissue dissociation. We thus focus on a newly developed exploratory tool called ClumpSeq, which involves sequencing small clusters of naturally attached cells via conventional single-cell sequencing protocols.</p>
<p>The focus of this thesis is computational and centered on developing deep learning models for analyzing ClumpSeq data. We validate our model on both simulated and publicly available data to show that it can extract cell type specific information from clump data. Based on these results we show that we can obtain novel biological insights that are otherwise not available from single-cell data.</p>
<p>Furthermore, we apply and adapt the ClumpSeq protocol to study the process of developmental hematopoiesis in the embryo, which is known to be influenced by the surrounding microenvironment. We generate a novel ClumpSeq dataset and show that our computational model provides insights about the composition of the niche surrounding the emerging hemogenic endothelium.</p>
|
|---|