Fig. 2: Modeling performance of Distance-AF. | Communications Biology

Fig. 2: Modeling performance of Distance-AF.

From: Distance-AF improves predicted protein structure models by AlphaFold2 with user-specified distance constraints

Fig. 2

The dataset of 25 targets was used. a comparison of Distance-AF(D-AF) against AF2 (blue circles), Rosetta (orange squares), and AlphaLink (green triangles) in terms of four metrics, RMSD, TM-Score, GDT-TS, and GDT-HA. b Modeling results with perturbated distance constraints. For a set of six distance constraints used for a target, the distance was perturbated randomly between ±5.0 Å for 10 times. Thus, 10 (perturbations) x 25 (models) =250 model results were plotted. Resulting models were evaluated in the four metrics relative to the average perturbation. c The model quality in the four metrics are shown with a different number of distance constraints. Four targets were used, 1MKMA, 2Z99A, 5HXYA, and 5KECA.The number of constraints used were 0, 1, 2, 3, 4, 6 (our default setting; shown in crosses), 10, 14, 18, 22, 26, and 30.

Back to article page