A platform for Mendelian randomisation using summary data from genome-wide association studies
All analyses, data extraction and more can be performed using the TwoSampleMR R package. Additionally, you can use the R package to analyse your own outcome datasets.
To begin analysis in the web application please review the data access agreement and accept by logging in with your google account.
Beta phase release
1.1.13 a4947c (21 June 2017)
0.1.2 (16 November 2016)
To use MR-Base using the TwoSampleMR R package directly:
See our sister website LD Hub for automated LD score regression:
Suggest new studies
If there are GWAS summary datasets you would like to see included in MR-Base, please could you enter the relevant details at the link below (but make sure the data is not already included in MR-Base).Click here to suggest new studies
Mendelian randomization using summary data from genome-wide association studies (GWAS) is an increasingly important tool for appraising causality in hypothesized exposure-outcome pathways. The approach can, however, be technically challenging and time consuming to implement. We have therefore created a new platform built on harmonised summary data from multiple GWAS called MR-Base that greatly simplifies the implementation of Mendelian randomization. In addition to simple lookup requests for individual SNPs across multiple GWAS, MR-Base automates implementation of two-sample Mendelian randomization, including effect allele harmonisation across separate studies, LD pruning to ensure independence of genetic variants and diagnostic and sensitivity analyses. See our methods paper for more details on the design and scope of MR Base. More general information on the principles, assumptions and limitations of Mendelian randomization can be found in the papers recommended on this page.
Intro to MR
Background on Mendelian randomisation
Davey Smith G, Ebrahim S. 'Mendelian randomization': can genetic epidemiology contribute to understanding environmental determinants of disease? Int J Epidemiol. 2003 Feb;32(1):1-22.
More background on Mendelian randomisation
Davey Smith G, Hemani G. Mendelian randomization: genetic anchors for causal inference in epidemiological studies. Hum Mol Genet. 2014 Sep 15;23(R1):R89-98
Haycock PC, Burgess S, Wade KH, Bowden J, Relton C, Davey Smith G. Best (but oft-forgotten) practices: the design, analysis, and interpretation of Mendelian randomization studies. Am J Clin Nutr. (in press)
Two-sample Mendelian randomization methods
Pierce BL, Burgess S. Efficient design for Mendelian randomization studies: subsample and 2-sample instrumental variable estimators. Am J Epidemiol. 2013 Oct 1;178(7):1177-84
Accounting for horizontal pleiotropy
GWAS studies, databases and consortia
We are grateful to the following GWAS studies, databases and consortia who have kindly made their summary data available:
- ADIPOGen (adiponectin)
- AMDGene (age related macular degeneration)
- Amundadottir et al (pancreatic cancer)
- Baranzini et al (multiple sclerosis)
- C4D (coronary heart disease)
- Capasso et al (Neuroblastoma)
- CARDIoGRAM (coronary heart disease)
- Ciampa et al (prostate cancer)
- CKDGen (chronic kidney disease and renal function)
- Deming et al (protein QTLs)
- DIAGRAM (type 2 diabetes)
- Duerr et al (Inflammatory bowel disease)
- EGG (birth and child anthropometrics)
- GABRIEL (asthma)
- GCAN (anorexia nervosa)
- GEFOS (bone)
- GIANT (adult anthropometrics)
- GLGC (lipids)
- GPC (personality)
- GTEx consortium (gene expression QTLs)
- GUGC (urate)
- HaemGen (Haemotological and platelet traits)
- Hom et al
- ICBP (blood pressure)
- IGAP (Alzheimer's disease)
- IIBDGC (Inflammatory bowel disease)
- ILCCO (lung cancer)
- IMSGC (multiple sclerosis)
- Kettunen et al (metabolites)
- Kiel et al (bone structure)
- Li et al (Alzheimer disease)
- Li et al (Upper gastrointestinal cancers)
- MAGIC (glycemic markers)
- Maraganore et al (Parkinson's Disease)
n et al (ischaemic stroke)
- MDACC (melanoma skin cancer)
- Mueller et al (diabetic nephropathy)
- NHGRI-EBI GWAS catalog
- Okada et al (Rheumatoid Arthritis)
- Olfson et al (alcohol dependence)
- Pankratz et al (Parkinson disease)
- PGC (psychiatric diseases)
- Rajaraman et al (glioma)
- ReproGen (Age at Menarche)
- Roederer et al (immune system)
- Shin et al (metabolites)
n-S nchez et al (Parkinson's Disease)
- Smith et al (bipolar disorder)
- SSGAC (education & cognitive performance)
- Stahl et al (rheumatoid arthritis)
- TAG (smoking)
The majority of data in MR Base is kindly made available for use by many research organisations and consortia .
These terms (Terms) set out the basis on which the University of Bristol, a body incorporated by Royal Charter under number RC000648 having its administrative offices at Senate House, Tyndall Avenue, Bristol BS8 1TH (University) agrees to provide you with access to the MR-Base platform (Platform) and through the Platform summary data from genome-wide association studies (GWAS Data).
1. Licence: The University grants you a non-exclusive, non-transferable revocable licence to access and use the Platform for private or non-commercial research purposes only. If you wish to access and use the Platform for commercial research purposes, please forward your enquiry to firstname.lastname@example.org.
2. Ownership of MR-base: Subject to paragraph 1, nothing in these Terms grants you any right to, or in, any intellectual property rights of any nature (whether existing now or in the future and whether registered or unregistered) in the Platform.
3. Downloading: You agree that you will not attempt to download GWAS Data from the Platform in bulk or otherwise use the Platform in a way that would or might adversely affect the performance or operation of the Platform for other users.
You agree to cite any use of the Platform in the form set out in the
5. Ownership of GWAS Data: GWAS Data may be protected by copyright, database rights and other intellectual property rights around the world. Unless otherwise stated in any notice accompanying any particular GWAS Data, all such rights are reserved by the contributor of the GWAS Data and you agree to observe and comply with any specific licence terms specified by such contributors.
6. Identification of data subjects: You agree not to use, combine, manipulate or transform the GWAS Data in any way that would or might enable you to identify any living individual to which the GWAS Data relates, in breach of data protection laws anywhere in the world.
We do not guarantee that (a) the Platform or any GWAS Data will always be available or interrupted; (b) the Platform or any GWAS Data will be accurate, complete, free from errors or omissions or secure or free from bugs or viruses; or (c) that the result of using the Platform or any GWAS Data will be accurate, adequate or fit for any particular purpose (more general information on the principles, assumptions and limitation of Mendelian randomization can be found on the papers recommended in the
8. Limitation of liability: You assume sole responsibility for the results obtained from the use of the Platform and any GWAS Data and for conclusions drawn from such use. The University shall have no liability for any damage or other loss whatsoever arising out of or in connection with your use of the Platform or any GWAS Data. The University shall not be liable in any circumstances whether in contract, tort (including for negligence), misrepresentation (whether innocent or negligent), restitution or otherwise for any special, indirect or consequential loss, costs, damages, charges or expenses however arising under these Terms including but not limited to loss of funding or loss of opportunity, goodwill or reputation. Nothing in these Terms excludes or limits any liability which cannot be excluded or limited by applicable law.
9. Suspension and termination: The University may, at its sole discretion, suspend, withdraw, discontinue or change all or any part of the Platform (in respect of any single user, group of users or all users of the Platform) without notice and whether or not arising from any breach of these Terms. The University will not be liable to you in such circumstances.
10. Changes to these Terms: The University may revise these Terms at any time by amending this page. Please check this page from time to time to take notice of any changes, as they will be legally binding on you.
11. Governing law and jurisdiction: These Terms and any claims or disputes arising out of or in connection with them (including non-contractual disputes) shall be governed by the laws of England and Wales whose courts shall have exclusive jurisdiction to settle the same.
Login button should appear below...
Choosing instruments for the exposure
To use two sample MR to estimate the causal effect of an exposure on an outcome, the first step is to identify SNPs that are robustly associated with the exposure. These summary statistics for these SNPs can be taken from a sample from which there is no data on the outcome.
Please provide instruments by choosing from one of the data sources below, or by uploading your own data. You can choose multiple exposures to be analysed, and multiple instruments per exposure.
Select outcomes for analysis
The MR Base database houses a large collection of summary statistic data from hundreds of GWAS studies. In order to perform two sample MR, the SNPs that were selected for the exposures will be extracted from the outcomes that you select here.
Please select the outcomes that you want to test for being causally influenced by the exposures.
Studies available in MR base
Most two sample MR methods require that the instruments do not have LD between them.
If a particular exposure SNP is not present in an outcome dataset, should proxy SNPs be used instead through LD tagging?
An important step in two sample MR is making sure that the effects of the SNPs on the exposure correspond to the same allele as their effects on the outcome. This is potentially difficult with palindromic SNPs.
Select methods for analysis
Many methods exist for performing two sample MR. Different methods have sensitivities to different potential issues, accommodate different scenarios, and vary in their statistical efficiency.
Once you have selected exposures, outcomes, and analysis options you are ready to perform the analysis.