Alcoholic liver disease (ALD) is a serious global health problem. It encompasses a spectrum of pathological conditions, ranging from simple hepatic steatosis, steatohepatitis, fibrosis, alcoholic hepatitis, to liver cirrhosis. Unfortunately, no definitive diagnostic markers exist for ALD (or its different phases), and diagnosis requires a liver biopsy which itself carries significant risk. As a result, management of ALD is frequently empiric. In recent years, some progress has been made using metabolomics to identify potential biomarkers of ALD in animal models and human cohorts. However, global metabolomic profiling of ALD in humans has proceeded slowly and as of today, no studies have been performed that relate metabolomic profiles with pathological changes occurring during the development of ALD. Our working hypothesis predicts that biomarkers specific to ALD may be more effectively identified by applying integrative machine learning to the analysis of data from two state-of-the-art analytical approaches, i.e., metabolomics and imaging mass spectrometry (IMS). As such, we propose to use plasma metabolomics (Specific Aim 1), and histological analysis and liver tissue IMS (Specific Aim 2) in three mouse models of ALD (alcohol-induced steatosis, hepatitis or mild fibrosis) to gain unique insights into the feasibility of using these approaches to identify pathogenic markers of ALD. Ethanol-induced damage to the liver results in alterations in cellular function that can be documented as changes in the metabolome of biological fluids (plasma) and hepatic cells. Metabolomics, the analysis of low molecular metabolites (e.g., lipids and small molecules) in a sample, can be used to directly investigate changes in biochemical pathways induced by alcohol in the liver, such as occurs during ALD. Tissue IMS maps molecules in a tissue section, thereby allowing the quantitation of lipids, proteins and metabolites within a tissue in unprecedented detail. When interfaced with histological analysis of a paired adjacent tissue section, the cellular source of the mapped molecules may be identified. We strongly believe that the integration of metabolomics, IMS and histology (Specific Aim 3) using integrative machine learning will greatly enhance our understanding of the biochemical basis of ALD pathophysiology, and in so doing, allow the development of diagnostic tools that can be used to detect biomarkers in other forms of ALD, thereby improving early diagnosis and treatment of ALD. The management and interpretation of large metabolomics and proteomic data generated as part of the project (10-100GB of raw IMS data per single tissue section) require advanced data-analytics solutions. We will capitalize on our recently published bespoke machine learning solution (?BASIS?) for interrogation of large ?-omics? data to identify metabolic/signaling pathways and their downstream metabolites disrupted in ALD. The novelty of this proposal relies on the use of cutting-edge approaches that will allow identification of novel biomarkers and their cellular sources in predictable animal models of ALD. Such information will form a basis for more effective diagnosis and prediction of the progression of ALD. Successful completion of the proposed studies will form a foundation upon which studies in human biological fluids will be conducted in the future. In addition, it is anticipated that our studies will also lay the foundation for examination of the molecular mechanisms associated with other forms of alcohol-induced tissue injury. Such knowledge will facilitate the development of more effective treatments of alcohol abuse.
Excessive alcohol consumption induces alcoholic liver disease (ALD). Unfortunately, no definitive diagnostic markers exist for ALD (or its different phases), and diagnosis requires a liver biopsy which itself carries significant risk. Because the early stages of ALD can potentially be reversed by sobriety, regular screening of the general population and early diagnosis are essential. The overarching goal of this application is to establish metabolomic analyses and tissue imaging mass spectrometry coupled with integrative machine learning to identify novel pathways in and biomarkers for ALD that will be applied in humans.
