Yes. Constructing PRS for diseases is hard (for reasons such as those mentioned by Isabel) - then here the added difficulty is that the PRS accuracy is measured across the population.

Hi Jeffrey, thanks for reading our paper. The power of a MMAA tool depends on the R2 of the Step 1 PRS (ie, proportion of phenotypic variance explained). For non-ascertained diseases, R2 is inevitably low (by contrast, if we had equal numbers of cases and controls, R2 would be much higher).