Mendelian Randomisation

Mendelian Randomisation analysis to explore the causal relationship between a potentially modifiable risk factor and an outcome. This script compacts the input and analysis steps in TwoSampleMR. It allows command line input of the instrumental data for 3 approaches - 1) top independent loci, 2) full summary statistics or 3) Using MRBase data.

The aim of this script is to allow fast, standardised analysis of numerous traits using pre-release and MRBase data.

Getting Started

Packages:

• optparse
• TwoSampleMR

Pre-processing:

If using approaches 1 or 2, you need to match format column names correctly. The column names must include SNP, effect_allele, other_allele, eaf, beta and pval. Optionally you can include ncase, ncontrol and samplesize (fully list available in TwoSampleMR documentation).

Column names can be edited with sed:

sed -i -e '1s/BP/pos/' -e '1s/A1/effect_allele/' -e '1s/A2/other_allele/' -e '1s/FRQ_A_67390/eaf/' -e '1s/OR/beta/' -e '1s/SE/se/' -e '1s/\tP\t/\tpval\t/' PGC3-cp.tsv

That says, on line 1 only, substitute BP with pos, A1 with effect_allele, etc. Note that '\t' is included for the 'P' column to avoid replacing 'SNP' with 'SNPval'. If you replaced OR with beta, remember that you need to use the -b flag to convert this accordingly.

Flags:

• -i/--instruments*: The approach being used for MR analysis. Depending on your desired approach, enter toploci, sumstats or mrbase.
• -s/--snps: The file location for exposure data. Applicable if -i is toploci or sumstats.
• -o/--outcome*: Method for outcome data - enter sumstats or mrbase.
• -f/--file: File location for sumstats of outcome. Applicable if -o is sumstats.
• -r/--r2: Cutoff for R2 to detect linkage disequilibrium and clump SNPs. Default value is 0.1 but R2 < 0.01, 0.001 have also been used in the literature.
• -e/--exp*: Name of the exposure.
• -d/--disorder*: Name of the outcome.
• -b/--beta: Required for sumstats and toploci only. Enter 'exposure', 'outcome' or 'both'.

*: compulsory flags.

Usage:

To perform MR using locally stored chronotype loci for exposure and MRBase data for schizophrenia as outcome:

./1_performMR.R  -i toploci -s ukb-chronotype-toploci.csv -o mrbase -e "chronotype" -d "schizophrenia"  

Note: The trait naming conventions can cause issues when using mrbase. Case sensitivity is applicable so for example "Depressive symptoms" will work but "depressive symptoms" will not. You can search your trait of interest at https://gwas.mrcieu.ac.uk/datasets/ to identify the naming convention used.

Use the search feature in IEU OpenGWAS at the link above to identify the naming convention for your trait, eg. chronotype:

Output:

Results are output into a results directory, in a subdirectory with the naming convention exp.(exposure).out.(outcome). The files outputted to this directory are:

Tables:

• heterogeneity.csv: Test for heterogeneity among causal estimates.
• mrresults.csv: Results for MR. MR results included are inverse-variance weighted (IVW), MR Egger regression and Penalised weighted median MR. These were selected to best encompass MR performed in previously reported MR studies.
• pleiotropy.csv: Performs MR Egger and returns intercept values.
• singlesnpMR.csv: Performs 2 sample MR on each individual SNP

Plots:

• scatterplot.png: Scatter plot of each MR analysis performed.
• funnelplot.png: Identify asymmetry which indicates directional pleiotropy.
• loo.png: To explore whether a single SNP drives causal associations.

Log:

• out.log: Log of stdout including flags selected and any command outputs. The radial MR results can also be found here as they print

Roadmap

The TwoSampleMR has a machine learning function that I would like to explore and intergrate into this script.

I would also like to integrate the package MRlap and eQTL data.

Potential Features

Limitations

Currently set-up is applicable for European samples. Top loci file must be in CSV format (the norm). Summary stats must be in TSV format (the norm)

Contact

Shane Crinion shanecrinion@gmail.com s.crinion1@nuigalway.ie

Reference

Hemani G, Zheng J, Elsworth B, Wade KH, Baird D, Haberland V, Laurin C, Burgess S, Bowden J, Langdon R, Tan VY, Yarmolinsky J, Shihab HA, Timpson NJ, Evans DM, Relton C, Martin RM, Davey Smith G, Gaunt TR, Haycock PC, The MR-Base Collaboration. The MR-Base platform supports systematic causal inference across the human phenome. eLife 2018;7:e34408. doi: 10.7554/eLife.34408