Applications of polygenic scores to individual disease prediction have so far been less successful, although proof of concept has been established through simulations [2]. Several studies have shown that a limited number of top ranking markers can discriminate disease cases from unaffected subjects, but the degree of discrimination falls short both of clinical utility and the maximum achievable from genetic data [16]–[18]. The use of a mass of markers across the whole genome has been explored, but to date has not yielded a noticeable improvement in discrimination [14], [19].
