| Summary: | Prion diseases are lethal neurodegenerative disorders caused by infectious proteins called prions. All known susceptibility variants in human prion disease are found in the prion protein gene (PRNP), but there is evidence that additional, non-PRNP susceptibility loci exist. Genome-wide association studies, an exome-sequencing study and an exome-array study have been conducted by the MRC Prion Unit in order to identify those loci. None of these studies have resulted in novel discoveries yet. Data integration could overcome the pitfalls of single-dataset analysis to discover novel susceptibility factors. In this project, Integrative Causal Analysis was adopted as a framework for data integration with the aim to identify all causal relationships between variants and prion disease that are consistent with all prion datasets and prior biological knowledge. Firstly, a theory of causal discovery from genetic datasets was formulated and causal discovery was applied to the datasets from the studies mentioned above. Secondly, an algorithm for causal meta-analysis of genetic datasets with overlapping sets of variants was designed and applied to a combination of the datasets in order to increase the power of learning causal relationships. Thirdly, a variant-filtering approach based on causal prior knowledge was devised as another method to increase power: Publicly available biological data and prior knowledge were integrated into a directed graph whose nodes comprise the disease and molecular entities in the cell and are associated with genomic regions, and whose edges denote causation. Candidate causal variants were subsequently identified from the ancestors (causes) of the disease in the graph. The application of the methods to prion disease resulted in a number of candidate susceptibility variants to be further investigated. The methods are also applicable to other diseases and have the potential to lead to novel discoveries in those diseases.
|
|---|
