Both physiological and psychiatric diseases are major health concerns affecting individuals worldwide. Currently, the advent of omics studies has provided researchers with an opportunity to investigate and identify biomarkers as a measurable indicator to predict disease risk. Several omics studies have identified markers for psychiatric disorders and physical diseases. However, improvement in omics data-generation technology does not guarantee the discovery of valid biomarkers; nor does discovered markers necessarily explain omics pathways of diseases. Hence, this study presents the principles and methods of multiomics analyses to overcome the current limitations of single omics marker-based disease risk prediction by compiling multiple omics results from two separate studies: stressomics and cardiomics. In the stressomics study, combined data from the methylome and transcriptome were used to develop machine-learning models to predict the risk of depression and suicide. The prediction model combining markers selected in accordance with loosely defined criteria presented a higher prediction accuracy than that of the markers selected with conventional criteria. Moreover, the classifier models displayed high prediction accuracies in predicting the appropriate labels the patients and controls. Further, the regression models developed for predicting the psychiatric scales were also successful. The second study, cardiomics, applied genomic data to predict the risk of acute myocardial infarction among young people. No variant was observed at the “genome-wide significance” level after performing a genome-wide association study. However, the polygenic risk score determined from the cumulative effect of whole-genome wide variants could distinguish patients with early-onset acute myocardial infarction and predict their subsequent cardiovascular events. Thus, disease risk prediction using integrated multiple markers can outperform the prediction by a single marker, even when each integrated marker has a low predictive accuracy by itself. Finally, this study elucidates the principles, methods, applications, and guidelines for multiomics analyses to help accelerate the utilization of omics data for future studies.
Publisher
Ulsan National Institute of Science and Technology (UNIST)