Table 2.

Improvement of protein secondary structure prediction with the addition of two "residue-specific" RSA thresholds, compared with this improvement using a fixed [4%, 16%] RSA threshold.

	Applied Threshold
	Fixed([4%,16%])	Mean ± standard deviation	Tertiles
A	-3.03	-0.83	-0.93
C	2.63	1.92	0.74
D	-0.95	1.70	1.50
E	-2.54	-1.02	3.98
F	0.71	10.14	9.44
G	0.90	9.28	7.94
H	-4.30	-2.14	-3.30
I	-7.49	-14.49	-15.16
K	10.37	26.73	13.31
L	3.04	4.14	2.87
M	-3.58	-5.45	-5.57
N	-1.21	2.46	-0.10
P	1.14	2.47	1.84
Q	0.53	0.07	-0.76
R	2.80	5.10	3.04
S	4.36	13.13	12.32
T	3.27	8.84	5.81
V	1.57	0.13	-5.72
W	-2.30	0.40	0.25
Y	4.17	9.73	10.30
Total Improvement	5.44	8.24	7.17

The values show percentage of improvement using leave-one-out cross-validation. The bold-underlined values are those values that show improvements when they are compared with the original GOR method. See the text for more details.