Comparing genome-wide significant and chemosensory variants as instruments for dietary patterns in Mendelian randomisation

Background: Diet is a modifiable risk factor for cardiometabolic disease, yet establishing causality remains challenging. Mendelian randomisation (MR) leverages genetic variants as instrumental variables (IVs) to enable causal inference. Method: Using two-sample MR, we assessed the causal effects of four principal component-derived dietary patterns (DPs) - Unhealthy, Healthy, Meat-based, Pescatarian - on twelve cardiometabolic outcomes: body mass index, coronary artery disease, high-density lipoprotein cholesterol, low-density lipoprotein cholesterol, total cholesterol, triglycerides, systolic blood pressure, diastolic blood pressure, type 2 diabetes, fasting glucose and insulin, and glycated haemoglobin. Two sets of IVs were employed: conventional genome-wide significant variants associated with each DP, rigorously filtered for pleiotropy and directionality; and biologically informed variants in chemosensory receptor genes, given the role of taste and smell perception in shaping food choices. Results: Using conventional IVs, the Pescatarian DP reduced fasting insulin ({beta}IVW = -0.10 pmolL-1 per SD increase in Pescatarian DP score, 95% Confidence interval [CI] [-0.15, -0.04]; P = 1.19x10-3), which survived multiple sensitivity analyses. Associations between the Unhealthy DP and elevated blood pressure and glycated haemoglobin were likely undermined by heterogeneity and pleiotropy, with insufficient IVs for robust sensitivity testing. Chemosensory receptors yielded null findings, reflecting insufficient power. Conclusion: Rigorously filtered conventional IVs supported the causal nature of well-established diet-disease relationships, demonstrating MR's utility in strengthening causal inference in nutritional epidemiology. Chemosensory IVs demonstrated limited utility for DPs, likely reflecting the heterogeneous and complex sensory profiles of overall diets. Future efforts should consider using guideline-based dietary scores to facilitate translation of findings.