Leveraging multi-omics for endotyping to identify subtypes and mechanisms of cardiometabolic diseases

Project Summary Prior studies have identified several risk factors related to cardiometabolic diseases. However, the rates of risk factor elevation are highly variable among patients. For example, despite obesity being identified as a primary risk factor for cardiometabolic diseases, a subgroup of obese patients does not develop downstream cardiometabolic complications. This suggests that there are other mediating mechanisms, independent of known risk factors, underlying cardiometabolic diseases. This project will leverage multi-omics data in conjunction with clustering approaches capitalizing on genome-wide association study (GWAS) data in BioVU, GWAS and other omics data generated by the NHLBI Trans-Omics for Precision Medicine (TOPMed) program and metabolomics data collected at 3 time points in a prospective cohort of bariatric surgery to identify: 1) clusters of individuals that represent cardiometabolic diseases (defined as novel/endotypic determinants of diseases), 2) endotypes of each cardiometabolic disease by clustering individuals within each disease (determined as pathobiological mechanisms related to disease) and 3) endotypes of responsiveness to bariatric surgery. To explore these overall study goals, the similarity network fusion method will be used along with the consensus clustering approach. Specifically, this project aims to 1) create genetically predicted levels of transcriptome, proteome and metabolome in 54,000 individuals in BioVU, use these predicted levels to construct clusters of individuals and examine whether these clusters represent diseases using the BioVU phenotype data, 2) construct clusters of individuals using direct measurements of multi-omics data (transcriptome, proteome and metabolome) in over 5,000 individuals in TOPMed, construct genetically predicted levels of the obtained clusters and impute them in BioVU using the BioVU GWAS data and explore whether these predicted levels represent cardiometabolic diseases using phenotype data in BioVU and 3) create metabolomic driven clusters using the plasma metabolomics data at baseline, 3 months and 12 months post bariatric surgery in 104 patients, explore the association of the identified clusters with cardiometabolic responsiveness and identify novel baseline metabolomic predictors of responsiveness to weight loss surgery . The proposed study will leverage multilayered –omics data using a novel and innovative network modeling analysis, providing Dr. Bagheri with critical skills, tools and experience to complete the research aims. This will be achieved by the accomplishment of the following two training goals which will help her to become an independent ‘big data’ cardiometabolic scientist: 1) gaining more in-depth knowledge of small molecule metabolism in biologically-informed cardiometabolic disease subtypes and 2) gaining expertise and expand her existing experience in bioinformatics and statistical multi-omics integration methods.