Skip to main content

Advertisement

Table 4 Performance comparison of QMEAN to other single model scoring functions based on the MOULDER test set.

From: QMEANclust: estimation of protein model quality by combining a composite scoring function with structural density information

Scoring function Mean ΔRMSD [Å] Std. dev. [Å]
torsion 4.50 4.06
pairwise Cbeta, SSE 1.48 3.00
Salvation Cbeta 1.06 1.91
SSE agreement 0.92 1.24
ACC_agreement 0.79 1.07
pairwise all-atom, SSE 0.68 0.96
QMEAN5 0.42 0.59
QMEAN6 0.40 0.59
SIFT 5.68 5.20
Anolea_Z 1.94 2.29
SOLVX 1.76 2.21
Xd 1.68 2.63
FRST 1.55 2.41
MP_SURF 1.36 1.90
MP_PAIR 1.20 1.70
EEF1 1.09 1.52
GB 1.06 1.36
DOPE_BB 0.96 1.27
PROSA_COMB 0.89 1.52
GA341 0.84 0.86
MODCHECK 0.83 1.29
MP_COMBI 0.82 1.19
DFIRE 0.81 1.37
DOPE_AA 0.77 1.21
ROSETTA 0.71 1.05
SVM_SCORE 0.46 0.66
  1. The table shows the RMSD difference (in Ångstrom) between the selected model by the scoring function and best model in the ensemble, averaged over the 20 protein targets of the MOULDER test set. In order to increase the robustness of the statistics, each calculation is repeated 2000 times on random subsets of 25% of the model ensemble. For comparison, the mean ΔRMSD and standard deviations for QMEANclust (based on consensus scoring of all 300 models) are 1.15 and 1.39 Å respectively. For a detailed comparison of QMEAN and QMEANclust see Table 5.