Fig. 6
Performance of eRankPPI, ZDOCK and ZRANK on the BM58 dataset. Ranking accuracy is assessed by the percentage of successful cases based on a-c iRMSD and d-f PCS. Each algorithm is evaluated against a, d experimental structures, as well as b, e high-quality and c, f moderate-quality protein models. Black dashed lines correspond to the upper bound estimated by taking the best of all 2000 models constructed for each target