8th Munich Metabolomics Symposium
Applications of clinical metabolomics in oncology and cardiovascular diseases
Virtual event November 12th, 2021
Part 2 - Cardiometabolic diseases
Organized by Helmholtz Zentrum Munich, Technical University Munich, and biocrates
Dr. Rui Wang-Sattler
Helmholtz Zentrum Munich, Germany
Abstract:
Large metabolomics datasets contain errors and noise. This renders the identification of true metabolite biomarkers challenging. Thus, providing efficient and reliable computational tools to extract clean and reliable information from metabolomics datasets is crucial to strengthen their application and utility in clinical research. To normalise metabolomics data, we have developed TIGER as a novel tool (released as an R package), which integrates the random forest algorithm into an adaptable ensemble learning architecture. By training with quality control (QC) samples and testing different QC and subject samples, TIGER enables the efficient removal of intra- and inter-batch technical variations. We demonstrate the utility of TIGER with three datasets constructed from targeted or untargeted metabolomics data. Furthermore, machine learning approaches (e.g. the supervised models with labelled input data) enable the early prediction of chronic kidney disease (CKD). We identify and evaluate prospective metabolite biomarkers and the best set of predictors of CKD in KORA (Cooperative Health Research in the Augsburg Region) individuals with prediabetes and type 2 diabetes. We further assess the organ-specific pathophysiology of the identified candidate biomarkers’ in two biofluids and six organs of leptin receptor-deficient mice and wild type controls. Animal studies indicate an additional cross-talk between adipose tissue, adrenal glands and liver that could contribute to circulatory accumulation of our identified metabolite biomarkers.
