CRC seminar: Hyungwon Choi, Ph.D.

Speaker

Hyungwon Choi, Ph.D., Depatment of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, China

Dr. Choi completed his PhD at the University of Michigan, Ann Arbor and is currently an Associate Professor in the Department of Medicine, Yong Loo Lin School of Medicine, National University of Singapore, and Chair of Cardiovascular – Metabolic Disease research program. His main research focuses on the development of computational methods to enable seamless application of mass spectrometry-based proteomics and metabolomics technology to biological and clinical studies. As Director of Singapore Lipidomics Incubator, he also leads analytical workflow development for untargeted and targeted metabolomics and lipidomics technologies with application to cardiometabolic diseases.   

Title

Partial correlation network as a vehicle to integrate high-dimensional omics data

Abstract

Network-based integration is a powerful technique to synthesize information from multiple sources of experimental data and understand important variations. However, if the underlying network needs to be estimated from the data directly, the estimation task can become daunting, especially in high-dimensional space such as omics data. In this talk, I will present a novel graphical model estimator ACCORD that enables the estimation task up to a million variables on high-performance computing environment. I will describe the method in laymen terms and demonstrate the method through two application studies. First, we underscore the importance of estimating gene co-expression network from transcriptomic data accounting for ultrahigh-dimensional epigenomic data (270,000 CpG islands) in TCGA liver cancer data, for robust prioritization of transcription regulatory factors. Second, we use the algorithm to identify correlated modules of circulating markers and echocardiographic imaging data and build predictive network model for major adverse cardiac events in a dual population study. In both cases, network-based data integration approach enables identification of key variables and their networks in the data that would otherwise be deprioritized.

Host

Professor Philipp Kaldis