Metabolomics enables the identification of putative biomarkers for application in many areas, including numerous diseases, healthy aging and even precision nutrition. Recent studies have shown some promise for early detection, therapy monitoring, disease recurrence and risk prediction, including our own work in these areas. While risk assessment and prediction approaches seem to work reasonably well, the development of translatable diagnostic tests has been slow to emerge. One major issue is the influence of many confounding factors such as genetics, environmental conditions, lifestyle, and even sample collection protocols that can affect metabolite levels. The resulting variability poses a major challenge for validating putative metabolite biomarkers for reliable clinical applications.
I will discuss some new approaches to alleviate the effects of these confounding factors on metabolite levels to improve biomarker performance and to assess sample data quality. In the first approach we are developing quantitative models of blood metabolite levels based on multisite cohorts to alleviate the often-seen site-to-site variation. A dramatic reduction (>90%) in this metabolite level variation was achieved based on modeling each metabolite using demographic and clinical factors and especially other metabolites. We have also identified some robust metabolite markers of sample quality, which can help uncover sample treatment history.
Finally, I will describe our efforts to identify reliable dietary intake biomarkers that can be used to correct highly variable participant recall data and provide more reliable disease risk predictions.