Table 3. Performance of random prediction method, in terms of MAE, on ANGLOR, SPINE X and ccPDB30 datasets for the prediction of phi and psi dihedral angle.

	Random PHI Prediction			Random PSI prediction
Residue/Dataset	ANGLOR	SPINE X	ccPDB30	ANGLOR	SPINE X	ccPDB30
ALA	40.4	34.3	33.7	83.6	82.3	83.0
CYS	44.6	42.7	42.2	88.5	88.7	88.3
ASP	47.8	42.2	41.5	84.8	83.2	83.6
GLU	40.3	33.2	33.3	83.6	78.9	80.8
PHE	43.9	40.6	40.5	88.0	89.5	89.4
GLY	88.5	87.8	88.2	87.3	88.1	88.2
HIS	49.2	46.7	46.7	89.6	88.7	87.1
ILE	34.9	32.9	32.5	88.1	88.5	88.1
LYS	44.0	38.1	38.4	85.9	84.4	85.6
LEU	34.3	30.5	30.2	87.9	85.4	86.2
MET	46.8	40.7	39.2	88.5	86.7	87.7
ASN	59.5	55.6	56.4	83.8	81.8	81.0
PRO	14.0	13.2	12.4	87.7	87.7	87.4
GLN	42.2	37.3	37.7	84.8	81.2	84.3
ARG	42.9	39.2	39.4	86.0	85.2	86.3
SER	49.7	42.8	42.1	89.7	89.9	89.7
THR	41.4	36.8	35.7	89.0	89.8	88.6
VAL	37.7	34.5	34.0	86.7	86.9	86.0
TRP	40.4	38.2	38.8	90.1	88.7	89.0
TYR	42.2	41.4	40.6	89.6	89.3	89.2
ALL	44.7	40.4	40.2	86.8	85.8	86.1
Helix (H)	36.3	32.0	32.0	82.7	78.4	80.5
Sheet (E)	44.9	44.2	43.2	90.7	93.7	92.2
Coil (C)	51.1	46.4	45.9	87.9	88.2	87.5