Table 1.

Comparison of the performance between the models trained individually and in an integrated manner.

Feature	AMAE (Individual)	(SD)	AMAE (Integrated)	(SD)	p-value
tASA:	18.73	0.92	17.43	0.75	< 2.2e-16
scASA:	24.32	0.79	22.32	0.96	< 2.2e-16
npASA	22.44	1.2	22.00	1.01	0.02
Phi:	12.16	0.77	12.13	0.74	0.82
Psi:	12.55	0.61	12.53	0.6	0.82
Kappa:	8.51	0.57	8.20	0.51	< 2.2e-16
Alpha:	10.21	0.64	10.23	0.63	0.84
HHC(AUC):	67.04	1.58	67.29	1.19	0.26

AMAE represents the average of the protein-wise mean absolute error (or AUC, the area under the ROC curve, in the case of HHC) and the standard deviation (SD) indicates the protein-wise variation of the AMAE values. The p-values are obtained by Welsh's t-test implemented in the R-programming language. The Integrated model outperformed the individually trained ones for the ASA and bend angle (kappa) predictions, but showed no statistically significant difference for the other conformational angles.