Step 7. Collect Information About the Current R Session.

Contact

The Centre for Artificial Intelligence Driven Drug Discovery (AIDD) at Macao Polytechnic University.

Location:

匯智樓 (WUI CHI)-4/F, N46B, Rua de Luís Gonzaga Gomes, Macau

Email:

kefengl@mpu.edu.mo

Call:

(+853) 8599 6883

Open Hours:

Mon-Fri: 10AM - 16PM


Two-step or Mediation Mendelian randomization

Mediation analysis seeks to explain the pathway(s) through which an exposure affects an outcome.

The main purpose of a two-step MR (mediation analysis) is to investigate whether a mediator can mediate the effect of exposure on the outcome. It is generally applicable for identifying potential mechanisms underlying the relationship from exposure to outcome.

Click here for a demo:



ALBERT
Exposure to Outcome (Step 1)

Causal Effect Estimation from Exposure to Outcome.

Identify significant SNPs from the GWAS summary data pertaining to exposure, and eliminate those in linkage disequilibrium. Subsequently, extract the remaining SNPs from the GWAS results associated with the outcome, taking care to ensure that these SNPs are not directly related to any confounders or the outcome itself. Finally, the causal effect of exposure on the outcome can be calculated (presumed to be beta0).

Exposure to Mediation (Step 3)

Causal Effect Estimation from Exposure to Mediation.

Identify significant SNPs from the exposure GWAS summary data, remove those with linkage disequilibrium, and then extract the remaining SNP information from the GWAS results for the mediator, ensuring that these SNPs are not directly related to confounders or the mediator itself. Finally, we can calculate the causal effect of the exposure on the mediator (assuming it to be beta1).

Mediation to Outcome (Step 4)

Causal Effect Estimation from Mediation to Outcome.

Identify significant SNPs from the GWAS summary data of the mediator, remove those with linkage disequilibrium, then extract the remaining SNP information from the GWAS results for the outcome, ensuring that these SNPs are not directly related to confounders or the outcome itself. Finally, we can calculate the causal effect of the mediator on the outcome (assuming it to be beta2).


Here we have three beta values (beta0, beta1, and beta2), and now let's briefly interpret their significance:

(1) If beta0, beta1, and beta2 are all significant, it indicates there is a causal association from exposure to outcome, and this association may be partially mediated by the mediator variable. Generally, we can consider beta1 * beta2 as the indirect effect from exposure to outcome, and beta0 - (beta1 * beta2) as the direct effect of exposure on the outcome. Then, we can statistically test the direct effect beta0 - (beta1 * beta2) to see if it significantly differs from 0. If this direct effect significantly differs from 0, it suggests the presence of this direct effect.

(2) If beta0 is not significant, but both beta1 and beta2 are significant, it indicates that the association from exposure to outcome is entirely mediated by the mediator variable.

(3) If beta0 is significant, but at least one of beta1 and beta2 is not significant, it suggests that there is no mediating effect of the mediator variable in the causal association from exposure to outcome.

   Choose exposure data.


                                

 Information for exposure

   Choose mediation data.


                              

 Information for mediation

   Choose outcome data.


                              

 Information for outcome

Step 4. Filter instruments.

   Step 4.1. Genetic variants significantly linked to the exposure/mediation/outcome factor.

   Step 4.2. Perform LD clumping.

   Step 4.3. Calculate R2 and F value.

Exposure

Mediation

Outcome

Step 6. Perform two-sample Mendelian randomization.

   Options for MR analysis.