Genetic risk prediction is an important and widely investigated topic because of its potential clinical application as well as its application to better understand the genetic architecture of complex traits (Chatterjee, Shi, & García-Closas, 2016). Many polygenic risk prediction methods have been developed and applied to complex traits. These include polygenic risk scores (PRS)(Chatterjee et al., 2013; Dudbridge, 2013; International Schizophrenia Consortium et al., 2009; Palla & Dudbridge, 2015; Shah et al., 2015; Shi et al., 2016; Stahl et al., 2012; Vilhjálmsson et al., 2015), which use summary association statistics as training data, and Best Linear Unbiased Predictor (BLUP) methods and their extensions (de los Campos, Gianola, & Allison, 2010; Golan & Rosset, 2014; Maier et al., 2015; Moser et al., 2015; Speed & Balding, 2014; Tucker et al., 2015; Weissbrod, Geiger, & Rosset, 2016; Zhou, Carbonetto, & Stephens, 2013), which require individual-level genotype and phenotype data.
