UC San Diego

Coronary heart disease (CHD) is a leading cause of death in the United States. Currently, the main method of risk assessment is carried out through established risk score algorithms by using traditional risk factors. These algorithms mainly focus on long-term prediction, with the limitation on assessing risk for younger adults. In recent years, with the advancement of serum nuclear magnetic resonance (NMR), more studies of using metabolites to predict CHD have merged. Assessing the risk with metabolites provides insights into the underlying molecular mechanisms of CHD. This thesis explores that possibility of using metabolites as the predictors and is aiming to understand how much prediction power that machine learning methods could bring in this prediction task.