Causal inference in large datasets using genetic instrumental variables: extending Mendelian randomization techniques. (360G-Wellcome-100114_Z_12_Z)

My key goal is the development of techniques for understanding the underlying causal mechanisms behind biological processes using information about genetic variants in population-based studies. In particular, I aim to find and validate causal risk factors and biological pathways responsible for cardiovascular disease (CVD). Methods development: I will develop methods for the extension of Mendelian randomization to the detailed analysis of complex causal networks between measured variables, a nd the integration of large-scale genetic data with data on exposure variables from multiple data sources. Applied analysis: This will involve the analysis of large-scale datasets on cardiovascular risk, including a case-cohort study in Europe and a case-control study in Pakistan each with more than 10,000 CVD cases. The breadth of the data facilitates detailed assessment of the assumptions on which Mendelian randomization relies. The aim is the categorization of risk factors by their causal role, including mediators in causal pathways, and the direction of causal association between different factors. Dissemination of methods: I will engage the epidemiological and wider medical community using illustrative papers, communicating the potential of the methods available to a wide audience. One aim is to develop a software package for investigators working in Mendelian randomization.