Current extensive genetic research into common complex diseases, especially with the completion of genome-wide association studies, is bringing to light many novel genetic risk loci. These new discoveries, along with previously known genetic risk variants, offer an important opportunity for researchers to improve health care. We describe a relatively fast method, using a Mann-Whitney U-statistic and based on the concept of the optimal receiver operating characteristic (ROC) curve, to build a test with many ideal properties (e.g., having the highest overall discriminative ability), to estimate its classification accuracy, and to determine the sample size required for the verification of this accuracy. The proposed predictive test is asymptotically more powerful than tests built on other existing methods, can handle a large number of genetic and environmental risk predictors, and allows for loci that interact. The method uses an efficient procedure to handle missing data. Simulations indicate that this method often performs better than the use of either classification and regression trees or simple allele counting. An application to Type 2 diabetes finds interactions that are replicated in an independent genome-wide association study.