Metabolomics enables the identification of putative biomarkers for numerous diseases including many cancers. A number of recent studies have shown promise for early detection, therapy monitoring, disease recurrence and risk prediction, including our own focusing on colon cancer. However, the influence of many confounding factors such as genetics, environmental conditions, and life style can affect metabolite levels and poses a major challenge in moving forward with such metabolites biomarkers for reliable clinical applications.
Advanced machine learning methods and quantitative metabolomics approaches can be used to identify and validate promising biomarker candidates as well as to evaluate and even model the effects of confounding factors with the goal of improving biomarker performance. Our results show that demographic covariates, such as gender, BMI, and smoking status, exhibit significant confounding effects on metabolite levels which can be modeled effectively. These results not only provide new insights into addressing the major issue of confounding effects in metabolomics analysis, but also shed light on issues related to establishing reliable biomarkers and the biological connections between them in a complex disease.
Finally, I will describe our efforts to apply metabolomics to identify and validate promising dietary biomarkers. Our goal is to develop blood based biomarker panels that quantitate macronutrient intake (carbohydrates, protein, total energy), which can be used to better provide estimates of diet-related disease risk, to inform improved dietary guidance and uncover diet related disease mechanisms.