Dietary effects on type 2 diabetes risk – Using metabolomics networks to elucidate the underlying mechanisms

健康的饮食可以预防2型糖尿病，但人们对其潜在机制知之甚少。该项目旨在确定调节饮食暴露对糖尿病风险影响的代谢途径。护士健康研究是一项前瞻性队列研究，包括121 700名护士。在哈佛大学进行的这项人口研究中，饮食习惯和疾病发病率在过去40年里不断进行评估。如今，新的代谢组学技术涵盖了血液样本中的广泛代谢物，并生成准确的新陈代谢快照。最近，代谢组学在护士健康研究中嵌套的病例对照样本中进行了评估，包括在研究过程中患上2型糖尿病的基线健康女性和没有疾病的对照组。如此详细的新陈代谢、饮食习惯和疾病发病率的纵向信息意味着在现实生活条件下观察糖尿病发展的饮食依赖机制的独特可能性。这项系统流行病学研究包括四个工作包。首先，将基于代谢物之间的相关性构建数据驱动的代谢组学网络。然后，代谢物将与2型糖尿病的风险联系起来。接下来，对于糖尿病相关的代谢物，将调查它们是否受到已知与2型糖尿病风险相关的饮食暴露的影响。代谢组学网络将用于区分对2型糖尿病风险的直接影响和仅反映与其他代谢物的相关性的间接影响。重复的测量和较长的跟踪时间意味着有可能建立变化模型，例如将饮食习惯的变化与随后代谢组学特征的变化联系起来。最后，将根据前三个工作包的结果生成假设，然后将使用现代因果推理工具进一步研究。基因变异将用于在孟德尔随机研究中验证代谢物和糖尿病风险之间的联系，并在性别-饮食相互作用分析中确定饮食对代谢的亚组特定影响。假设性饮食干预对代谢和2型糖尿病风险的影响将在重复测量的基础上进行预测。预期结果将支持在人群水平上预防2型糖尿病的饮食干预，并将提供个性化的饮食建议。

A healthful diet protects against type 2 diabetes but the underlying mechanisms are poorly understood. This project aims to identify metabolic pathways that mediate the effects of dietary exposures on diabetes risk.The Nurses‘ Health Study is a prospective cohort study including 121 700 nurses. In this population study conducted by the Harvard University, dietary habits and disease incidence were continuously assessed over the last 40 years. Nowadays, new metabolomics techniques cover a broad spectrum of metabolites in blood specimens and generate accurate snapshots of metabolism. Recently, metabolomics profiles were assessed in a case-control sample nested within the Nurses‘ Health Study, including baseline-healthy women who developed type 2 diabetes in course of the study and a control group that remained disease-free. Such detailed longitudinal information on metabolism, dietary habits and disease incidence implies the unique possibility to observe diet-dependent mechanisms of diabetes development under real-life conditions.This systems epidemiology study comprises four work packages. First, data-driven metabolomics networks will be constructed based on the correlation between metabolites. Then, metabolites will be linked to type 2 diabetes risk. Next, for diabetes-related metabolites, it will be investigated whether they are influenced by dietary exposures that are known to be related to type 2 diabetes risk. The metabolomics networks will be used to differentiate between direct effects on type 2 diabetes risk and indirect effects that merely reflect the correlation with other metabolites. The repeated measurements and the long follow-up time imply the possibility to model change, e.g. relating change in dietary habits to subsequent change in metabolomics profiles. Finally, hypotheses will be generated based on the results of the first three work packages and will then be further investigated with modern causal inference tools. Genetic variants will be used to validate the links between metabolites and diabetes risk in Mendelian randomization studies, and to identify subgroup-specific effects of diet on metabolism in gen-diet interaction analyses. Effects of hypothetical dietary interventions on metabolism and type 2 diabetes risk will be predicted based on the repeated measurements.The expected results will support dietary interventions to prevent type 2 diabetes on the population level and will inform personalized dietary recommendations.