IndepMR: An Ontology-Driven, Ancestry-Aware Interactive R Shiny Framework for Automated Two-Sample and Multivariable Mendelian Randomization with Mediation Analysis Using Public GWAS Summary Data

<title>Abstract</title> <p>Background Mendelian Randomization (MR) leverages genetic variants as instrumental variables to estimate causal effects of modifiable exposures on health outcomes using publicly available genome-wide association study (GWAS) summary data. Despite growing accessibility of GWAS repositories (e.g. NHGRI-EBI GWAS Catalog, IEU OpenGWAS), a rigorous end-to-end MR workflow remains inaccessible to most applied researchers. There is no existing tool that provides automated ancestry matching, ontology-driven trait retrieval, and an integrated pipeline spanning two-sample MR, multivariable MR, and mediation analysis in one interface. Results We present IndepMR, an interactive R Shiny application that automates the MR workflow from study selection to mediation analysis. IndepMR introduces three contributions: (1) an ontology-driven trait search engine using Experimental Factor Ontology lineage paths, retrieving studies beyond keyword matching; (2) an ancestry independence engine classifying population descriptors into a five-group continental hierarchy with subgroup resolution, handling multi-ancestry populations via umbrella-term detection; and (3) an integrated five-tab workflow supporting two-sample MR, multivariable MR, and mediation analysis with automated instrument extraction, linkage disequilibrium clumping, allele harmonisation, sensitivity diagnostics, and exportable results. We applied IndepMR to the Body Mass Index (BMI) to Type 2 Diabetes Mellitus (T2DM) pathway in two ancestry groups. In the European pair, IndepMR replicated a significant causal effect of BMI on T2DM risk (IVW OR = 1.60, 95% CI 1.06 to 2.41, p = 0.026). In the Asian pair, it identified severe heterogeneity and pleiotropy alongside a direction-reversed estimate (IVW OR = 0.69, p = 0.045). A further demonstration using a non-independent mediator shows how the ancestry engine flags sample overlap that would bias the estimate. Conclusions IndepMR provides a reproducible, user-guided environment for MR analysis that enforces ancestry-matching criteria, prevents multi-ancestry overlap errors undetected by existing tools, and makes rigorous causal inference accessible to non-programmers. It is freely available as an open-source R code.</p>