Metabolomics has established itself as a powerful platform for identification of disease biomarkers and for providing insights into biochemical mechanisms of disease. Despite recent advances, key challenges to exploiting the full potential of metabolomics remain, including identification of the full range of metabolites found by untargeted analyses, and assigning them to metabolic pathways and networks. Our group has developed a framework and a set of tools for interactive pathway analysis and visualization of metabolomics data. While these and other knowledge-based tools proved to be useful, the scope of their application is limited to only a subset of known compounds. A large portion of known metabolites including lipids, secondary metabolites and dietary compounds cannot be mapped to canonical metabolic pathways. With this in mind we recently developed new Debiased Sparse Partial Correlation (DSPC) modeling method for estimating metabolic networks that allows delineating direct interactions between metabolites from experimental data in settings where the number of features (metabolites) far exceeds the number of available samples. Despite the ability of DSPC to recover partial correlation networks in low dimensional settings, the small sample sizes available in many metabolomics studies still constrain the number of significant relationships between metabolites (i.e. network edges) that can be reliably estimated. To overcome this problem we propose to incorporate information from biochemical knowledge and existing metabolomics data sets into our data-driven DSPC modeling method. To demonstrate the feasibility of our approach we will focus on type 2 diabetes, which has been extensively studied using metabolomics. We will use publicly available data to further develop and train the DSPC algorithm and apply it to the analysis of untargeted blood metabolomics profiles derived from women who participated in the longitudinal multi-ethnic Study of Women's Health Across the Nation (SWAN).
The specific aims of the project are:
Aim 1 : Extend the DSPC method to incorporate prior information about metabolites to enhance its capabilities and accuracy.
Aim 2. Use the enhanced DSPC method to analyze SWAN metaboomics data. Successful completion of this project will provide new data analysis methodology and tools that will enhance the value of metabolomics data by integrating existing biological knowledge and data to better estimate network structure. These tools will be broadly applicable across different disease areas and help to fill important gaps in analysis and interpretation of large-scale metabolomics data.

Public Health Relevance

Large scale measurement of small molecule metabolite levels (metabolomics) can provide biomedical researchers with a comprehensive view of normal and disease physiology. In this proposal, we will develop new computational methods that will help researchers to analyze metabolomics data in order to gain new insights into the mechanisms underlying the development of diabetes.

National Institute of Health (NIH)
National Cancer Institute (NCI)
Small Research Grants (R03)
Project #
Application #
Study Section
Special Emphasis Panel (ZRG1-BST-U (50)R)
Program Officer
Spalholz, Barbara A
Project Start
Project End
Budget Start
Budget End
Support Year
Fiscal Year
Total Cost
Indirect Cost
University of Michigan Ann Arbor
Ann Arbor
United States
Zip Code
Puskarich, Michael A; Evans, Charles R; Karnovsky, Alla et al. (2018) Septic Shock Nonsurvivors Have Persistently Elevated Acylcarnitines Following Carnitine Supplementation. Shock 49:412-419
Basu, Sumanta; Duren, William; Evans, Charles R et al. (2017) Sparse network modeling and metscape-based visualization methods for the analysis of large-scale metabolomics data. Bioinformatics 33:1545-1553