We applied the same methods to predict T2D in Latino target samples from the SIGMA T2D data set (row 3 of Table 1). We used publicly available European summary statistics from DIAGRAM(Morris et al., 2012) (12,171 cases and 56,862 controls; effective sample size = 4/(1/Ncase + 1/Ncontrol) = 40,101) as European training data and SIGMA T2D genotypes and phenotypes (SIGMA Type 2 Diabetes Consortium et al., 2014) (3,848 cases and 4,366 controls; effective sample size = 8,181) as Latino training and validation data, employing 10-fold cross-validation.
