Figure 5

The best TM-score observed among the top ranked protein pairs. Protein pairs in the fold level dataset was ranked by their FCC and for each pair (of a query and a template) the structural similarity of the structural models of the query inferred by the alignment was compared with the native structure of the query using TM-align. At each rank in the x-axis, the best TM-score was plotted. For protein pairs with equal FCC, TM-scores are averaged among the pairs. To determine corresponding residues in protein pairs, TM-align (the left column), HHpred (the middle column), and SUPRB (the right column) were used. Three different bases for residue contact definitions are used, the Cα-Cα distance (the first row), the Cβ-Cβ distance (the second row), and the heavy atom distance (the third row). The panels are ordered in the same way as Figure1