From Mendelian Randomization to Hypothesis Free Causal Inference

Funder: Medical Research Council

Dimensions: grant.3560505

Research Organisations


Obtaining reliable evidence on causes of disease and influences on disease progression is problematic, since many other factors than the ones under study may influence the outcomes. Using genetic variants that mimic potentially modifiable factors that may cause disease or influence its course has become a widely-used approach to improving such understanding. Known as “Mendelian randomization” the methodology is still under development. This programme of work will contribute to strengthening the reliability of the approach, and, in particular, investigate the application to identifying treatments for existing disease. Technical Summary Scientific abstract Background: The MRC Integrative Epidemiology Unit (MRC-IEU) has pioneered the Mendelian randomization (MR) approach, which is now showing exponential growth in terms of breadth of application and number of publications. This has been fuelled by the coalescence of four trends (1) the rapid increase in the number of identified genetic variants related to exposures of interest to clinical investigators; (2) the maturation of UK Biobank and other large-scale population studies with accessible data, including from genome wide association study (GWAS) analysis; (3) the expansion in available data from GWAS consortia; and (4) the popularization of 2-sample MR analyses (2SMR), which allow causal estimates to be made from summary level data. We have developed a range of new methods that allow stronger inference to be drawn from MR studies and released a platform (MR-Base) for high-throughput 2SMR analysis. Both have been rapidly taken up by the scientific community. Our aims will considerably expand the range of questions that MR can address, include: Aim 1: To increase the clinical utility of MR by conducting studies of disease progression, which will provide MR evidence that speaks directly to disease treatment Aim 2: To develop approaches that subject the MR assumptions to further scrutiny through negative controls and the assessment of pleiotropy, together with developing structural equation modelling as a flexible strategy for analysis of MR studies Aim 3: To conduct MR studies in family based data, which provides additional robustness checks, and allows estimation of maternal and paternal influences on offspring outcomes Methods: The three aims will be advanced in parallel, within a programme that will utilise simulation, substantive analysis, replication and triangulation of evidence. Translation will be facilitated by the new NIHR Bristol Biomedical Research Centre (Davey Smith Scientific Director) and through expanding our collaborations with industry.

Resulting publications

Funding information

Funding period: 2013-2018

Funding amount: EUR 2449427

Grant number: MC_UU_12013/1

Research Categories

Main Subject Area

Fields of Research