Is cisMRcML MR or colocalization?

[jielab]

Hi, there:

I found that the least significant P-value of your example data mr_dat$exp_df is 4.2308e-09. Does this means that all SNPs included in exp_df must be genome-wide significant?

Previously I thought that cisMRcML is like a coloc analysis, where non-significant SNPs are included. So, should I run coloc after running cisMRcML?

Thanks! JH

[ZhaotongL]

Hi, there:

I found that the least significant P-value of your example data mr_dat$exp_df is 4.2308e-09. Does this means that all SNPs included in exp_df must be genome-wide significant?

Previously I thought that cisMRcML is like a coloc analysis, where non-significant SNPs are included. So, should I run coloc after running cisMRcML?

Thanks! JH

Hi Jie,

Thank you for your interest!

cisMR-cML is an MR method, which uses cis-SNPs around the gene of interest as instrumental variables. It is not a colocalization method. In our manuscript, we ran colocalization after cisMR to evaluate our findings from the cisMR analysis.

[jielab]

Thanks, Dr. Lin!

cisMR-cML only includes SNPs whose P <5E-08, while colocalization needs to include all significant and non-significant SNPs, correct? For cis-pQTL, usually there are not many SNPs with P <5E-08, do you think it is better to change this threshold to something like P <5E-05?

Your paper said that we identified three potential drug targets, PCSK9, COLEC11 and FGFR1 for CAD. But the biggest drug target for CAD is HMG-CoA reductase. Somehow your analysis could not pick up this protein?

Your colocalization analysis only provided causal evidence for PCSK9, COLEC11, but not for FGFR1. What is the potential reason for this? Frankly, I have no confidence with colocalization analysis, even though every paper reports colocalization analysis after running MR. For example, colocalization analysis with eQTL data is actually fake data, because the eQTL is imputed from GTeX and no eQTL of study participants was actually measured.

Best regards, Jie