The Genome-wide MVMR app runs a genome-wide multivariable Mendelian randomisation (MVMR) workflow.
MVMR evaluates the association of multiple exposures with an outcome in the same model. This allows you to estimate the association of each exposure with the outcome while accounting for the other selected exposures.
Use this app when exposures may be biologically related or correlated, and you want to assess their independent associations with an outcome.
Before you begin
You will need:
- One or more outcome GWAS studies
- Two or more exposure GWAS studies for a multivariable analysis
- A destination project for the job and output files
Name your job
Enter a clear Prefix to identify the analysis. You may also add an optional Suffix to distinguish related runs.
[Screenshot 1: Highlight the Prefix and Suffix fields in the Job Identification section.]
For example:
- Prefix:
Lipids_to_CAD_MVMR - Suffix:
GWsig_r001
Select outcome study or studies
Use Outcome Study ID to select the GWAS study or studies representing the outcome of interest.
[Screenshot 2: Red box around the Outcome Study ID selector.]
The outcome is the trait, disease, or phenotype for which you want to evaluate the associations of the selected exposures.
Select exposure study or studies
Use Exposure Study ID to select the GWAS studies representing the exposures of interest.
[Screenshot 3: Red box around the Exposure Study ID selector.]
Select the exposures you want to include in the same multivariable model. MVMR is most useful when you need to account for potential overlap or correlation between exposures.
For example, you may include related lipid traits, body-composition measures, or molecular biomarkers as separate exposures.
Set the P-value threshold
Use P Value Threshold to define the statistical-significance threshold for selecting genetic instruments. The default is 5e-8.
[Screenshot 4: Red box around the P Value Threshold field.]
The default retains genome-wide significant variants. A more stringent threshold retains fewer instruments, while a less stringent threshold may include more candidate instruments.
Set the LD threshold
Use R2 Threshold to define the maximum LD allowed between selected variants. The default is 0.001.
[Screenshot 5: Red box around the R2 Threshold field.]
A lower R2 threshold retains more independent genetic instruments. The default of 0.001 is stringent and is commonly used for genome-wide MR analyses.
Select a project
Choose a destination project in the project_id dropdown. This is required.
[Screenshot 6: Red box around the project_id dropdown and the note directing users to Research Hub → Projects.]
If a suitable project is not available, create one in Research Hub → Projects before launching the analysis.
Launch the workflow
Review the selected outcome, exposures, instrument-selection thresholds, and project. Then select Launch App.
[Screenshot 7: Red box and arrow pointing to “Launch App”.]
When complete, access the output from the selected project or the job results area.
Interpreting results
The MVMR results estimate the association of each selected exposure with the outcome after accounting for the other exposures in the model.
Interpret these estimates in the context of:
- Instrument strength and number of available variants
- Correlation between exposures
- Consistency of GWAS populations and ancestry
- The assumptions required for Mendelian randomisation
MVMR can help distinguish potentially independent exposure–outcome associations, but results should be evaluated alongside biological knowledge, sensitivity analyses, and other supporting evidence.
Tips
- Use at least two exposures; this is what distinguishes MVMR from standard univariable MR.
- Choose exposures that are relevant to the same biological or clinical question.
- Use a clear job name when comparing different exposure sets or thresholds.
- Retain the default P-value and R2 thresholds unless your analysis plan specifies otherwise.