Research

The utility of metabolome and lipidome-wide profiling data can be elevated via integration with complementary modalities such as quantitative genomic, epigenomic, transcriptomic and proteomic profiles of metabolic enzymes, transporters and other signaling molecules modulating the metabolic activity. We develop advanced statistical computing algorithms to identify biochemical pathways and novel relationships between small molecules and gene products from multi-omic data. This work is enabled through innovative algorithms for graph inference from ultrahigh-dimensional multi-omic data and subsequent reconciliation with experimentally validated biochemical pathways. 

• Lee et al., Learning Massive-scale Partial Correlation Networks in Clinical Multi-Omics Studies with HP-ACCORD. Annals of Applied Statistics 2025; 19(4): 2759-2781.  
• Koh et al., An integrated signature of extracellular matrix proteins and a diastolic function imaging parameter predicts post-MI long-term outcomes. Frontiers in Cardiovascular Medicine 2023; 10: 1123682.