Novel graph theoretic biological pathway network analytics methods for analyzing and discovering Alzheimer's disease related genes

Abstract

BackgroundAlzheimers disease (AD) is the most common form of dementia among older people. It is a complex disease and the genetics and environmental factors behind it are not conclusive yet. Traditional statistical analyses are inadequate to identify variants, genes, or pathways capable of explaining AD as a unit. In this context, pathway network analysis based on a set of curated AD-specific genes identified in the literature can elucidate biological mechanisms underneath AD. Through the network, we can infer influential pathways that can together explain AD. Consequently, we can target those pathways and corresponding genes for further analysis to develop new drugs, discover novel AD-related genes, combine multiple hypotheses, and so forth. MethodsWe have developed a novel graph theoretic algorithm that can elucidate complex biology from a given set of disease-related genes. It constructs a weighted network of enriched pathways where similarity score between a pair of pathways is defined in a context-specific manner. To make the network robust, we employ topological overlap techniques on top of the raw similarity measure. We then provide the importance of each pathway with respect to the entire network, functional modules and importance of each pathway in a specific module, gene clusters, and so forth. We also provide a method to identify a set of novel genes that can further explain the disease-related genes and the disease itself. ResultsWe have employed our algorithms onto a set of AD-specific genes. It identified three distinct functional modules that are related to metabolism, cancer, and infectious disease related pathways. These findings are matched with three recognized hypotheses in Alzheimers disease, e.g. "metabolism hypothesis," "cell cycle hypothesis," and "infectious disease hypothesis." By analyzing the curated genes common among those functional modules, we can attain more understanding about this fateful disease. We have also identified 24 novel AD-related genes of which at least 14 genes are known to be involved in AD. ConclusionsWe developed a computational framework for analyzing biological pathways in a context-specific manner. It can be used in any sets of disease-related genes. We manifest its efficacy, reliability, and accuracy by employing a set of AD-specific genes.