Excess weight increases the risk of several diseases including cardiovascular disease, type 2 diabetes, kidney disease and various cancers. There is a need for preventative strategies for obesity-associated disease, especially for people in the overweight and moderately obese ranges where pharmacological intervention may not be suitable. Low-carbohydrate (ketogenic) diets are popular for weight control. Ketogenic diets increase circulating ketones, which can have favourable effects on cardiometabolic health markers. However, the ketogenic diet has a nutrient composition associated with harms (high-saturated fat/red meat, and low-fibre). The net effects of ketogenic diets on long-term health are unclear. Ketone supplements can increase circulating ketones and could provide benefits of ketosis without needing to adhere to a potentially harmful diet. Establishing causality between complex exposures (e.g., diet) and long-term outcomes (e.g., disease), is challenging. The MRC \& NIHR Review of Nutrition and Human Health Research (2017) highlighted an overreliance (as opposed to reasonable reliance) on observational studies as a key barrier to progression in the field of nutrition and health. Randomised controlled trials (RCTs) facilitate causal inference, but for long-term outcomes are expensive, time-consuming, and often suffer from waning adherence. Mendelian randomization (MR) can estimate causal effects subject to key assumptions. A challenge to these assumptions includes complex behavioural exposures (e.g., diet), which could be intercorrelated with causal factors. Our proposal will address these limitations with a novel combination of study designs to establish causal effects of ketosis (via diet and supplementation) on obesity-associated disease risk in humans. The investigators will combine a tightly controlled, short-term RCT, with MR to link short-term responses to long-term endpoints. The investigators will examine the circulating (blood) and tissue-specific (adipose) transcriptomic and proteomic responses in the fasted and postprandial state in response to our dietary interventions and translate these to MR by identifying single-nucleotide polymorphisms from genome wide association studies. This approach overcomes limitations of RCTs and MR, as adherence to diets will be confirmed with controlled feeding, and intermediate molecular traits as exposure for MR are less likely to be intercorrelated with causal traits.