Cross Validation Approach. A) 5-fold cross-validation split of the full sample data. B) Explanation of how 5-fold cross-validation is incorporated into the MDR method. First the sample data is randomly split into 5 intervals with representative numbers of cases and controls in each interval. MDR is performed on each of the 5 possible splits. Cross-validation consistency (CVC) is calculated for each of the best models and a final model is chosen which maximizes CVC and minimizes prediction error (PE).