PROJECT SUMMARY Metabolomics offers a comprehensive analysis of thousands of small molecules in biological samples. It can play an indispensable role in the growing systems biology approaches to unravel the relationships between metabolites and diseases. Liquid chromatography coupled to mass spectrometry (LC-MS) and gas chromatography coupled to mass spectrometry (GC-MS) have been used for high-throughput analysis of thousands of metabolites. However, the potential values of many disease-associated metabolites discovered by using these platforms have been inadequately explored in systems biology approaches for biomarker discovery due to lack of computational tools and resources to: (1) accurately determine the identity of most of the metabolites; (2) investigate the rewiring interactions among the metabolites due to diseases; and (3) integrate metabolite profiles with those from other omics studies to evaluate the relationships between the metabolites and the diseases at the systems level. Partly due to these limitations, poor generalizability of previously identified metabolite biomarker candidates has been observed, especially when they are evaluated through independent platforms and validation sets. Therefore, new methods are sought to find more generalizable metabolite biomarker candidates. The goal of this research program is to fill the gaps in metabolite identification and multi- omics integration by using systems metabolomics approaches that will enhance the role of metabolomics in systems biology approaches for biomarker discovery. Specifically, the proposed research program will utilize multiple resources (biological databases, spectral libraries, etc.) and innovative statistical, machine learning, and network-based methods for: (1) developing a comprehensive workflow for ranking putative metabolite IDs; (2) differential analysis of metabolite profiles based on changes in the levels of individual metabolites and pairwise interactions in disease vs. control groups; and (3) integration of metabolomics data with genomics, transcriptomics, proteomics, and glycoproteomics data to identify highly promising metabolite biomarker candidates. Our recent progress has led to acquisition of multi-omics data and development of computational tools for metabolite identification and integrative analysis. The performance of the proposed metabolite identification workflow in ranking putative metabolite IDs will be evaluated through experimental methods using reference compounds. The differential and integrative analysis methods will be used for selection of candidate biomarkers via multi-omics data acquired in biomarker discovery studies. The selected candidates will be evaluated by targeted quantitation using independent samples and platforms compared to those used for discovery. The outcomes of these experimental evaluations will be used not only to help refine the computational methods but also to identify promising biomarker candidates. In summary, the proposed research program seeks to capitalize on the power of network modeling, machine learning, and multi-omics data integration to improve the ability to find disease biomarkers that are likely to succeed in future large-scale biomarker validation studies.

PROJECT NARRATIVE The goal of this research program is to develop computational methods that enhance the role of metabolomics in systems biology approaches for biomarker discovery. Specifically, statistical, network-based, and machine learning methods will be developed for ranking putative metabolite IDs and selection of metabolite biomarker candidates. The performance of the proposed methods for metabolite identification and biomarker selection will be evaluated through experimental methods including: (1) comparing the fragmentation patterns of unknown analytes against those from reference compounds; and (2) targeted quantitative analysis of selected metabolites. 2