reference population, although method can be applied using other reference populations, such as the TOPMed reference population on the Michigan Imputation Server19. Since TOPMed samples are used as controls in GAWMerge, there will be sample overlap between the input data and the TOPMed reference population, which may cause bias and must be applied cautiously. Further work is needed to evaluate the bias of such an imputation strategy. The application of GAWMerge can be conducted beyond the phenotypes tested here, where we plan to expand its applications to other studies, such as opioid addiction in the near future.
