subset of top ranking markers. Two related but distinct applications of this score are then possible. Firstly, testing for association between the score and the trait in the replication sample can determine whether associated markers reside within those contributing to the score. Secondly and perhaps more usefully, the polygenic score can be used to predict individual trait values or risks of disease [2], potentially giving a predictor with better discrimination properties than one based on established markers only. Different considerations apply for these two applications, as the size of the replication sample has a direct bearing on the power of association testing, whereas the accuracy of individual predictions depends only on the size of the training sample.
