Computational methods and analyses to dissect the pathogenesis of Frontotemporal Dementia

Abstract

rontotemporal Dementia (FTD) is a devastating neurodegenerative disorder that typi cally manifests before the age of 65 and is characterized by progressive degeneration of frontal and temporal lobes as well as behavioural changes and problems with speech. Although great advancements in our understanding of FTD have been made during the last decades, there still does not exist a treatment that halts the progression of this dis ease. It is therefore necessary to further advance our understanding of the molecular mechanisms that cause FTD and that drive disease progression forward. In this thesis, I have contributed to the field of FTD research through the development of computational methods and the analysis of multi-omics datasets in order the develop new hypotheses for disease mechanisms in FTD. Studying complex tissues such as the human brain requires to carefully consider the contributions of diverse components of such systems. A major factor in transcriptomic experiments is cell type composition, as every cell has a unique transcriptional profile. In chapter 2, we have developed a deep learning-based cell type deconvolution algorithm that outperforms other methods and, importantly, also works well on post-mortem human brain tissue. The algorithm was rigorously tested and made available to the community as an open source, accessible python package and as web application. In chapter 3, we have analysed multi-omics datasets from post-mortem human brain tissue of FTD patients with mutations in the genes GRN, MAPT and C9orf72. Using an integrative data analysis approach, we could identify common and distinct affected path ways in these three genetic FTD subtypes. We leveraged the rich multi-omics datasets to identify new aspects of the disease, such as vulnerable neurons and increasing blood vessel percentages. In-depth analysis could highlight several regulator molecules, such as micro RNAs and transcription factors, that are likely to play important roles in FTD and therefore depict promising subjects for future studies. In chapter 4, we have performed co-expression module analysis of transcriptional data from seven different brain regions of patients with genetic subtypes of FTD. Using this comprehensive dataset, we have highlighted regions that are transcriptionally affected in different FTD subtypes and regions that seem not to suffer from the disease. We further more highlighted region- and disease-specific co-expression modules and pinpointed hub genes of potentially important function for these modules. Our analysis is the first that evaluates transcriptional deregulation at such diversity in FTD, and therefore provides ixAbstract valuable novel insights for the field of FTD.