PRS vs Mendelian Randomisation

Hi everyone, I was wondering aside from the methodological difference what the major theory difference would be in using a PRS vs MR approach to understand exposure outcome association. For example if we take the example of peripheral inflammation (as assessed by some biomarker) on BMI, what is the difference between a 2 sample MR using GWAS of peripheral inflammation and GWAS of BMI, vs. a PRS derived from the GWAS of peripheral inflammation applied to predict BMI in the target sample?


Hi Salah,

Using a PRS will improve the instrument strength, this is desirable in behavioural outcomes, when instruments are numerous and weak. The disadvantage is that then you are potentially introducing a lot more (horizontal) pleiotropy.