Effect sizes (wk,g) in model (1) can be estimated using multiple approaches. In this paper we compare penalized approaches such as LASSO (Least Absolute Shrinkage and Selection Operator)18 and the elastic net19 as well as the more naive simple polygenic score estimates. However, other statistical machine learning approaches42, such as Random Forest43 or OmicKriging44, can be used within the PrediXcan framework to develop prediction models.
