Although the main focus of complex disease genetics is to gain insights about disease biology, genetics can also be leveraged to build predictions of disease risk, which may become clinically useful as sample sizes increase73,74. A landmark study of schizophrenia showed that polygenic risk scores, constructed by summing the predicted effects of all markers below a P-value threshold in the training sample, produced predictions of schizophrenia risk in validation samples that were significantly better than random, and far more accurate than those based on the single genome-wide significant locus identified in the study75. This provided an early demonstration of the advantages of incorporating markers that do not attain genome-wide significance into polygenic risk scores to improve prediction accuracy for polygenic traits. An important issue in computing polygenic risk scores is that of LD between markers, which has historically been addressed by LD-pruning—either without regard to P-values75, or via informed LD-pruning76 (clumping) that preferentially retains markers with more significant P-values. More recent work has shown that explicitly modeling LD using an LD reference panel and estimating posterior mean causal effect sizes
