Table 3. Overview of the tables and graphs included in the MR-Base platform.

Tab	Details
MR results	A table with the causal estimates resulting from each MR method that was implemented. See Table 4 for an example based on use case 2. Estimates are presented in the units of the exposure SNP(s). Estimates are beta coefficients for the outcome and should be exponentiated if the unit of the outcome was a log odds ratio. P-values are calculated using a t-distribution.
Heterogeneity statistics	A table with statistics indicating the variation in the causal estimate across SNPs, i.e. heterogeneity. Lower heterogeneity indicates better reliability of results.
Causal direction test	The results of a test that uses variation explained in both the exposure and outcome to assess whether the direction of the results is likely to be correct. Note the test cannot determine whether a causal association exists.
Horizontal pleiotropy	The Egger regression intercept with its standard error and a p-value.
Single SNP analysis	A summary graph showing the individual effects of SNPs, calculated using the Wald ratio, along with the overall results to assess the consistency across SNPs. See Figure 2 for an example.
Method comparison plot	A graphical representation of the results given under the ‘MR results’ tab. This graph shows the effect of the SNP(s) on exposure against the effect of the SNP(s) on the outcome. The graph is structured so that the effect of the SNP(s) on the exposure is always positive and the effect of the SNP(s) on the outcome is directed accordingly. See Figure 3 for an example based on use case 2.
Leave-one-out analysis	A graph showing the results of MR analyses using the inverse variance weighted method when leaving one SNP out each time. This analysis can be used to assess whether the SNPs are consistent in terms of their effect on the overall outcome or whether the results are being driven by a single outlying SNP. See Figure 4 for an example based on use case 2.
Funnel plot	A graph to visually assess heterogeneity, particularly horizontal pleiotropy. Horizontal pleiotropy is likely if points are spread. Directional horizontal pleiotropy may be present if the graph is not symmetrical. See Figure 5 for an example based on use case 2.