dc.description.abstract | Treatment of paediatric cancer can result in late effects. Late effects are health problems that occur after cancer treatment has ended. One drug commonly used during treatment of paediatric cancers is dexamethasone. Dexamethasone has an anti-inflammatory effect. However, prolonged exposure to dexamethasone can lead to various health problems, including impaired bone development. The effects of dexamethasone treatment have been well studied in adult bone but not yet in young bone. There might be substantial differences in the effects of dexamethasone treatment between adult and young bone. To study this effect in young bone, young mice were treated with various dosages of dexamethasone. Hereafter, single-cell RNA sequencing was used on the tissue samples of these mice. Single-cell RNA sequencing is a technique used to capture and sequence individual cells. This results in cell-specific expression profiles. The expression profiles of dexamethasone treated and untreated cells can be compared to identify cell type specific effects of dexamethasone treatment. However, single-cell RNA sequencing data contains doublets. Doublets are technical artifacts which form when two cells are captured together. The doublet formation depends on the number of cells used during sequencing. Too many cells were loaded during the sequencing of the mice tissue samples, hence this data contains many doublets. Doublets cannot be recognized easily and compromise the data analysis. Therefore, they have to be detected and removed. DoubletFinder and scDblFinder are two existing doublet detection methods. These methods create artificial doublets by randomly picking and combining two real cells into one. Then these methods look at the gene expression of cells and identify the differences between the artificial doublets and the real cells. Cells that are closely related to the artificial doublets are likely to be doublets and hence annotated as doublet. DoubletFinder and scDblFinder differ in their machine learning approach to annotate doublets as well as in the generation of artificial doublets. scDblFinder sums and reweights cells, whereas DoubletFinder averages cells. The aim of this study was to evaluate the performance of DoubletFinder and scDblFinder. The best approach could then be applied to clean the single-cell RNA sequencing data. This study identified scDblFinder to perform better and faster than DoubletFinder on existing datasets for which an experimental ground truth reference was available. Hereafter, the single-cell RNA sequencing data was cleaned of doublets and the effect of dexamethasone was investigated. This was done by first annotating the cell types and then looking at the differences in expression profiles between the same cell types in treated and untreated cells. Annotating the cell types was done by comparing the expression profiles to a reference dataset for which the cell types are known and then importing the known labels of similar expression profiles to the dataset. This study reports that dexamethasone negatively affects several cell types, including chondrocytes and osteoblasts, these cells are found within the bone and contribute to bone formation. In adult mice dexamethasone has been reported to cause both a decrease in bone formation and an increase in bone resorption. The latter has not been identified in this study and could be a key difference between adult and young bone. | |