Diet is a major risk factors for chronic disease, yet efforts to measure and monitor diet are hampered by a reliance on self-reporting methods which suffer from large measurement errors. Dietary biomarkers, molecules from food that can be measured in blood or urine, can be used as objective measures of specific food intakes and has the potential to overcome these obstacles. However, only a few valid biomarkers exist today. Over the past 10 years, advancements in metabolomics and other technologies has provided enormous potential for large-scale untargeted screening of dietary biomarkers for e.g. epidemiological studies ultimately aiming for improved population health. In this project, we will use data on dietary exposures as well as plasma metabolomics, already collected and available, from a type 2 diabetes case/control study (n=600 cases & 600 individually matched controls) nested within the Västerbotten Intervention Programme Cohort. Using an R-based data analytical pipeline already established based primarily on multivariate predictive methodology (applied in e.g. SNIC 2015/6-60), we will identify biomarkers of individual dietary exposures (such as milk, coffee, fruit, vegetables, red meat) as well as dietary patterns established both from a priori defined patterns (such as the Healthy Nordic Diet) or from unsupervised factor analysis. In this prospective research design, dietary biomarkers (both individual markers and multivariate marker panels) will then be tested for associations with risk for future development of type 2 diabetes using conditional logistic regression (from baseline sampling) of GEE modelling (on baseline and 10 year follow-up samples). This is a continuation of SNIC 2015/6-60 (DiBiDi: Dietary Biomarker Discovery), but migrated to SENS, since data analysed are from human origin (GDPR).