Linking Human-Genetic and Host-Microbiome Associations to Molecular Mechanisms Using Protein-Protein Interaction Networks.

다국어 초록 (Multilingual Abstract)

We continue to better understand human disease through the study of human genetics and the human microbiome. Both fields use cohort study design, where the genes or microbiome communities of patients with a given condition are compared to a group of healthy controls. Systematically performing these comparisons to derive gene-disease associations and microbiome-disease associations is increasingly commonplace, but generating reasonable hypotheses for further study is still a challenge. Overcoming this challenge is vital to understand the causal molecular mechanisms of human disease.Quality-control and prioritization pipelines for human genetic variants from large-scale studies are hard to build, often involving several computational tools, databases, and tuning parameters. In Chapter 2, I present GeMSTONE, an online variant prioritization tool that allows researchers to leverage a large variety of resources to replicate and customize these pipelines without any computational experience or overhead. Understanding the location of disease-associated variants relative to protein interaction interfaces can help us understand whether they are likely to disrupt a protein interaction, thereby implicating a discrete molecular phenotype with the disease. In Chapter 3, I present Interactome INSIDER, an online resource that expands our network of protein interaction interfaces and performs widespread annotation of disease variants in this context.Determining the mechanisms behind host-microbiome disease- associations is particularly challenging, as there are few functional annotations for commensal microbiome proteins as well as sparse microbe- human protein interaction networks not involving extreme pathogens. In Chapter 4, I build a human-bacteria protein-protein interaction network that is used to detect the differential targeting of human proteins by commensal bacteria in association with disease.This thesis presents three different examples of mechanistic hypothesis generation from large-scale association studies. I demonstrate how annotation of variants using biological databases, structural interaction networks, and bacteria-human protein interactions can expand our understanding of the likely actors in human disease.