A brief comparison of polygenic risk scores and Mendelian randomisation

BMC Medical Genomics

Table 1 A non-exhaustive list of potential applications of PRS and MR

Applications	Polygenic risk score analysis	Mendelian Randomisation	Potential for bias due to horizontal pleiotropy?	Requires individual level data?^a
Identifying an effect of an exposure on an outcome	✓	✓	✓	MR = NO PRS = YES
Comparing outcomes in high versus low genetic risk for an exposure	✓	X	✓	YES
Examining gene x environment interactions	✓	✓	✓	YES^b
Prediction modelling	✓	X	X	YES
Quantifying shared genetic aetiology	✓	X	X	NO
Identifying downstream effects of liability to a disease	✓	✓ (“reverse MR” [3])	✓	MR = NO PRS = YES
Identifying biomarkers of disease	✓	✓	✓	MR = NO PRS = YES

^aWhere this column is marked ‘NO’, this indicates that the research question can likely be addressed using summary level data (from GWAS) instead
^bFor identifying gene environment interactions using MR, this can in theory be conducted with summary data provided that both the exposure and the outcome GWASs have been performed within the subgroups of interest. For example, if we examine the effect of BMI on dementia risk in APOE4 versus APOE3 carriers, we would need the GWAS of BMI and the GWAS of dementia to be performed separately in APOE4 carriers and APOE3 carriers. However, in practise, most summary level data for both our exposure and outcome of interest are not available within subgroups, so we usually require individual level data to examine gene x environment interactions

ISSN: 1755-8794