Table 3.

Coefficients of correlation showing the interdependence between the prediction performances for various structural features.

	MAE (tASA)	MAE (scASA)	MAE (npASA)	MAE (Phi)	MAE (Psi)	MAE (Kappa)	MAE (Alpha)	AUC (HHC)
MAE (tASA)	1	0.93	0.88	0.09	0.1	0.1	0.11	0.19
MAE (scASA)	0.93	1	0.93	0.07	0.08	0.07	0.06	0.21
MAE (npASA)	0.88	0.93	1	0.07	0.08	0.07	0.07	0.2
MAE (Phi)	0.09	0.07	0.07	1	0.46	0.35	0.42	-0.03
MAE (Psi)	0.1	0.08	0.08	0.46	1	0.41	0.47	-0.05
MAE (Kappa)	0.1	0.07	0.07	0.35	0.41	1	0.76	-0.05
MAE (Alpha)	0.11	0.06	0.07	0.42	0.47	0.76	1	-0.04
AUC (HHC)	0.19	0.21	0.2	-0.03	-0.05	-0.05	-0.04	1

The prediction performance for the ASA features is highly correlated with each other. The prediction performance for some conformational angles (e.g., alpha and kappa) is well-correlated, whereas others (e.g., Phi and Kappa) are only weakly correlated.

MAE: Mean absolute error (% ASA or degrees), AUC: Area under the ROC curve; HHC: helix-helix contact.