Untargeted metabolomics and data analysis strategies to discover biomarkers

Time

20 December 2018, 13:00

Place

Auditorium A1-01.01, Festauditoriet, Bülowsvej 17, Frederiksberg.

Opponents

Associate Professor Faidon Magkos (chair), Department of Nutrition, Exercise and Sports, University of Copenhagen, Denmark

Professor Rikard Landberg, Food and Nutrition Science, Chalmers University of Technology, Sweden

Professor Hanne Christine S. Bertram, Department of Food Science, Aarhus University, Denmark

Supervisor

Professor Lars Dragsted, Department of Nutrition, Exercise and Sports, University of Copenhagen, Denmark

Co-supervisor

Associate Professor Timothy Ebbels, Department of Surgery & Cancer, Faculty of Medicine Imperial College London, United Kingdom

About the thesis

Biomarkers are a useful tool to measure exposures and their biological effects and to evaluate system susceptibility. They have been widely applied in the nutrition area to evaluate dietary intake and its resulting biological consequences and the nutritional status of the study subjects. However, the existing classification schemes of biomarkers are ambiguous leading to uncertainty about their application. With the development of metabolomics technology, the biomarker research field is experiencing rapid changes and new biomarkers are continuously discovered. Challenges also emerge as metabolomics data is large scale, noisy and complex. Therefore, efficient data analysis tools are needed to extract the most relevant information from it.

In this thesis, a classification scheme for the currently applied and newly discovered dietary and health biomarkers in the nutrition field is developed, the data analysis strategy in untargeted metabolomics studies is optimized for the more precise discovery of biomarkers and applied for the discovery of biomarkers related to onion consumption.

We developed an improved scheme for biomarker classification based on their intended use rather than the technology or outcomes and suggested six subclasses. For the discovery of exposure biomarkers, different bi- and trilinear PLS models for variable selection in untargeted metabolomic studies were compared. Bilinear PLS model with group × time-response as dummy Y showed the best performance. This model has been applied in an onion intervention study and eight biomarkers of onion intake have been successfully discovered.

Read more: https://nexs.ku.dk/english/calendar/2018/phd_qian-gao/