Fig. 3From: PUMAS: fine-tuning polygenic risk scores with GWAS summary statisticsModel-tuning performance on real GWAS data. A PUMAS performance on the EA training set. B Prediction performance on two validation sets for EA. C PUMAS performance on the AD training set. D Prediction performance on two validation sets for AD. The X-axis shows the log-transformed p-value cutoffs in PRS which is the tuning parameter of interest. The Y-axis indicates predictive R2. EA educational attainment, AD Alzheimer’s diseaseBack to article page