Developing a Molecular Taxonomy for Immune-Mediated Diseases

About This Grant

ABSTRACT: One of the most fundamental limitations in medicine is reliance on clinical phenotype as the primary way to diagnose and classify disease. For inflammatory diseases, clinical phenotype alone is not sufficient to describe underlying disease pathogenesis. For example, cytokines like interleukin (IL)-6, tumor necrosis factor (TNF)-α, and IL-17A can all drive arthritis, but clinical features such as arthritis or rash may not reflect the causal cytokine. As a result, there are no precision biomarkers to guide targeted cytokine therapy: phenotypic classification systems do not stratify patients by causative mechanisms that guide therapy. The lack of precision medicine strategies is one of the most significant unmet needs in the field, leading to poor outcomes for over 40 million patients with inflammatory diseases. This includes >1 million patients with a phenotype-driven diagnosis of “rheumatoid arthritis”: 40-50% of whom fail first biologic therapy, and 10% with disease refractory to multiple consecutive targeted therapies. We propose addressing this problem by combining expertise in immunogenetics and translational immunology (Dr. Schwartz); systems immunology and machine learning (Dr. Das); and statistical genetics (Dr. Wang).We will leverage another key finding: shared molecular mechanisms like IL-6 signaling can drive different phenotypically defined diseases like arthritis, vasculitis, and scleroderma. Using machine-learning approaches, we will build a phenotype-blind molecular taxonomy for immune diseases. Drawing from the success of phenotype-blind tumor genotyping in cancer therapy, we will adapt machine learning approaches developed in the Das lab to build novel tools (SIDER, Significant Interaction Directional Effect Representation; MAGen Multiscale network Approach for Genetic architecture). We will use SIDER and MAGen to construct an “omnigenic model”, connecting a small number of disease-causal core genes to many peripheral genes through complex biological networks. We will ground our approach in human biology using Dr. Schwartz’s expertise to build on >500 core genes that cause monogenic inborn errors of immunity (IEI). We will use interpretable machine-learning and network approaches developed by Dr. Das to develop molecular modules that will be applied to large population databases leveraging Dr. Wang’s expertise We will investigate two aims: (1) understanding monogenic/oligogenic disease architecture and (2) developing novel molecular endotypes for common diseases. This approach will demonstrate a molecular taxonomy is a relevant and actionable approach that will transform the identification, study, and treatment of immune diseases and beyond.