The outlook for disease and trait prediction is more challenging. To date the severe shortfall in the accuracy of genetic predictors has generally been ascribed to incomplete coverage of marker panels or failure to identify sufficiently many associated markers. Here, however, no criteria for declaring individual significance are imposed, but neither does the calculation force the predictor to include markers that contribute no information. Under this pragmatic approach it results that tens of thousands of subjects, at least, are needed to derive predictors that are clinically useful. Furthermore, previous results on the potential accuracy of genetic prediction [17], [20]–[22] only become relevant at very large sample sizes. Such numbers are now coming within reach of national biobank projects and international consortia, so the emergence of useful genetic predictors may not be too far off, although such large samples create issues of effect heterogeneity that are not addressed here. Recent estimates of the proportion of markers having effects also suggest that the more optimistic scenarios shown in Table 4, Table 5, Table 6 may apply [9], [33]. Although the focus here
