Evaluation of the effectiveness of mixing and matching initial models. The accuracies of the initial comparative predictions and the refined CF models are evaluated with the CαRMSD measurements of models for 40 CASP7 targets [see Additional file 1: Table 1]. For each target, the CαRMSD of the best scoring CF model is compared and plotted against that of the best scoring initial model. The accuracies of the best scoring CF models are consistently higher than or equal to those of the best scoring initial models. Mixing and matching between two initial models (A) leads to an average improvement of 0.2 Å CαRMSD between the best scoring CF models and the best scoring initial models. Mixing and matching between multiple models (B) leads to an average improvement of 0.4 Å CαRMSD. The targets with the best improvements for the two methods are 0.7 Å and 1.6 Å CαRMSD, respectively.