Furthermore, by using information from a GWAS of a second related trait or disease, it is possible to calculate a PRS in a population which has been ascertained with a particular trait or disease in mind. An example would be the use of alcohol consumption GWAS summary statistics (that is, examining units of alcohol consumed per week) in order to characterize risk for alcohol use disorders, which are related constructs but not directly linked. If the genetic causes of those two traits overlap and the predictive performance of the PRS is good, then significant association with the related PRS will also be observed; however, the SNPs providing maximal distinction between case and control groups in the related trait may differ from those which would maximally distinguish the primary trait. Hence, it is important to identify and understand the primary GWAS which was used as the basis to generate the PRS before making a judgment on the accuracy (or lack thereof) of a PRS for a specific analysis or trait. In the example above, alcohol consumption turns out to have only partial genetic overlap with risk for alcohol use disorder and thus the accuracy of prediction is limited 23.
