Hi everyone,
I’m currently working on a Mendelian Randomization (MR) analysis, and I’m a beginner in this field. My goal is to investigate the association between two diseases — heart failure and type 2 diabetes.
Here’s my workflow so far:
I downloaded GWAS summary statistics for heart failure and type 2 diabetes from the FinnGen database.
I used eQTL data from the GTEx v8 dataset (aorta tissue) as the exposure.
I performed clumping on the eQTL data using PLINK with the following parameters:--clump-p1 5e-8 --clump-r2 0.01 --clump-kb 10000
In R, I filtered the original eQTL data according to the clumped results, keeping only variants with p < 1e-5.
Then, I used the two GWAS datasets as outcomes and the filtered eQTL dataset as the exposure to perform separate MR analyses for the two diseases.
After obtaining the MR results, I filtered them again by p-values and took the intersection of significant SNPs from the two analyses.
Finally, using this intersected set of SNPs, I opened a 100 kb window around each SNP in both GWAS datasets and the eQTL data, and performed colocalization (coloc) analyses for each disease separately.
I then took the intersection of the two coloc results as well.
However, I didn’t obtain any overlapping results after this process, which is quite frustrating. Since I haven’t received formal training in this area, I’m not sure whether my pipeline has major flaws. I’d really appreciate it if someone could help me identify possible issues. If my explanation isn’t clear enough, I can share my R script for review.
