Table 1.
Dataset used in our study containing 16 protein complexes.
Table 2.
Methods used for comparison in study with a short summary of their approach and scoring function.
Table 3.
Statistical measures used to test the performance of each method in predicting ΔΔG values.
Fig 1.
Calculated ΔΔG values (x-axis) compared to experimental ΔΔG values (y-axis) for each method tested in this study.
Black, red, and blue lines are simple linear regressions from which r are derived. The red points are a scatter for Ab complexes and the blue points are for non-Ab complexes. The dashed line is the y = x line measuring perfect agreement between predicted and experimental ΔΔG values. The solid black, red, and blue lines indicate a linear relationship between calculated and experimental observations for all data points, Ab complexes, and non-Ab complexes respectively. The top values in black, red, and blue match the root-mean-square error and the bottom values indicate r for all values, Ab values, and non-Ab values respectively.
Fig 2.
Performance of each method for non-Ab complexes (401 total mutations) in predicting true ΔΔG values (⍴c), linearly correlated ΔΔG values (r), and rank order (⍴ and τ).
The error for each method is reported under the correlation points.
Fig 3.
Receiver operating characteristic (ROC) curves for non-Ab complexes of the classification of variants as stabilizing (ΔΔG < -0.5 kcal/mol) or destabilizing (ΔΔG > 0.5 kcal/mol).
The values in the legend represent the area-under-curve (AUC). The higher the value, the better method is at discriminating between destabilizing and destabilizing mutations.
Table 4.
All methods r with respect to certain subsets.
Fig 4.
Performance of each evaluated method for Ab complexes (253 total mutations) in predicting true ΔΔG values (⍴c), linearly correlated ΔΔG values (r), and rank order (⍴ and τ).
The error for each method is reported under the correlation points.
Fig 5.
Receiver operating characteristic curves of the classification of variants that are more destabilized or less destabilized than 0.5 kcal/mol.
The values in the legend represent the area-under-curve (AUC). The higher the value, the better the prediction capability of the method.