Mendelian Randomization, Liftover

Question

[lag]

Hi,

I'm about to do several two-sample mendelian randomisation analysis. I'm in doubt if it is important that both exposure and outcome dataset are from the same assembly genome (hg19/hg38). Also, I'm running several MR with the same outcome data, but different exposure datasets - is it important that every analysis are on the same build or is not important as the analysis are using rsIDs?

Thank you in advance for any input to this.