Skip to content

Chemosensory Mendelian Randomization Pipeline

Summary

The Chemosensory Mendelian Randomization (MR) Pipeline is a biologically informed framework designed to select valid genetic instruments (instrumental variables) for dietary exposures. Developed by Hwang et al. (as of 2026-06-01), this multi-stage filtering pipeline prioritizes nonsynonymous variants in taste and olfactory receptor genes to study diet-disease relationships. It minimizes standard MR biases like horizontal pleiotropy and reverse causation by ensuring genetic instruments are anchored in primary sensory biology rather than health-status-associated dietary reports.

Method and Filtering Stages

The pipeline selects and refines instruments through the following sequential gates: 1. Variant Prioritization: Selects nonsynonymous variants (missense, nonsense, or frameshift mutations with minor allele frequency $\ge 1\%$) in taste receptor type 1 (TAS1R), type 2 (TAS2R), and olfactory receptor (OR) gene families. 2. Cohort Replication: Validates food-liking associations identified in a discovery cohort (such as the UK Biobank) in an independent younger cohort (such as the Avon Longitudinal Study of Parents and Children (ALSPAC)) to exclude age-related health confounding. 3. Liking-Intake Concordance: Confirms that candidate genetic instruments affect both food preference (liking) and actual food consumption (intake) in the same direction. 4. Socioeconomic Control: Excludes variants showing significant associations with socioeconomic status (e.g., Townsend Deprivation Index) to prevent confounding. 5. Food Specificity & Linkage Disequilibrium (LD): Filters out variants associated with multiple foods to control for pleiotropy, taking into account LD ($r^2 \ge 0.1$) and co-consumption patterns. 6. Causal Directionality: Applies Steiger Filtering to ensure that the genetic variants explain more variance in the exposure (food liking) than in the disease outcome, preventing reverse causation by confirming the direction of causal flow towards the outcome as conceptualized in standard Mendelian Randomization.

Applications and Code Repositories

The pipeline was applied to identify 24 genetic instruments for 20 foods, including a high-confidence instrument for onion liking (variant rs6587467 in the OR2T6 gene) that was subsequently used to demonstrate causal associations between onion preference and lower blood pressure and reduced risk of type 2 diabetes.

Code for processing UK Biobank food liking and ALSPAC replication data is available in the public repositories maintained by the authors: - UK Biobank Pipeline: GitHub Repository - ALSPAC Analysis: GitHub Repository

Citations

  • Hwang, L. D., Lin, C., Evans, D. M., Martin, N. G., Reed, R. D., & Joseph, P. V. (2026). A biologically informed framework for instrument selection in dietary Mendelian randomization using chemosensory genetics. BMC Medicine, 24:359. DOI: 10.1186/s12916-026-04966-x. Source paper: s12916-026-04966-x.pdf
  • Sanderson, E. et al. (2026). Challenges and future directions for Mendelian randomization. Nature Genetics, 58(5), 984–994. DOI: 10.1038/s41588-026-02546-6. Source paper: s41588-026-02546-6.pdf