Predicting Cardiovascular Outcomes Using Diabetes-Induced Transcriptomic Networks - ABSTRACT Type 2 diabetes mellitus (T2DM) is an increasingly prevalent chronic disease that affects more than 400 million people worldwide. One of the major complications of T2DM is exacerbated atherosclerotic cardiovascular disease (CVD). Even when modern lipid and glucose control strategies are applied, T2DM is associated with a two- to four-fold increase in CVD risk, suggesting the effect of additional pathologies, such as inflammation. However, current tools to predict CVD outcomes for T2DM patients incorporate only clinical and demographic variables into their models, and they thus attain only a moderate ability to discriminate the highest-risk patients in need of targeted clinical intervention. Our lab recently discovered that monocyte-derived foam cells, which are well-known to play a central role in atherosclerotic CVD, can undergo both homeostatic (non-inflammatory) and pathogenic (inflammatory) foaming. Using a transcriptomic signature from pathogenic foam cells, our lab developed a CVD prediction model called CR30 which outperformed existing tools. To address the critical knowledge gap of identifying CVD risk specifically in T2DM patients, I analyzed monocyte transcriptomic data from the Multi-Ethnic Study on Atherosclerosis (MESA). From this preliminary analysis, I identified a transcriptomic signature unique to T2DM patients with CVD, containing a super-network downstream of the co- regulator proteins SNW1, NCOR2, and CITED2. We hypothesize that this transcriptomic super-network represents a unique molecular signature which can be used to improve prediction of atherosclerotic cardiovascular events in individuals with T2DM. In this proposal, I will test this hypothesis by applying two different strategies to develop predictive models. In Aim 1, I will apply supervised machine learning approaches to select a set of genes from my preliminary analysis which are predictive of T2DM-CVD outcomes. I will then test several modeling strategies in training and building a T2DM-CVD prediction model incorporating this gene set combined with clinical data. In Aim 2, I will use another approach to incorporate T2DM-CVD molecular signature into modeling by focusing on the transcriptomic super-network. I will generate enrichment scores for the super-network, then incorporate the scores as variables into model development. The long-term goal of this project is to identify biological risk factors for CVD in patients with T2DM. The anticipated impacts are the identification of novel targets for mechanistic studies and the advancement of biology-informed approaches to clinical outcomes prediction. The training goals of this proposal will provide me with biologically-informed quantitative skills. This interdisciplinary, highly translational project will leverage the innovative environment and unique opportunities in the sponsor’s lab and the University of Connecticut School of Medicine. The expected outcomes from this project will promote my career goals of becoming a next-generation physician-scientist capable of integrating biological knowledge and quantitative skills to solve clinical problems for patients with chronic disease.
