Table 2.

Comparison of results for several alphabets with the two architectures specified in Table 1, using SAM-T06 multiple alignments

Alphabet	bits saved		Qn
	IDaaHR	IDGaaH13	IDaaHR	IDGaaH13
str2	1.05	1.12	0.54	0.56
dssp	0.92	0.98	0.63	0.64
stride	0.89	0.94	0.66	0.67
bys	0.70	0.83	0.58	0.59
dssp-ehl2	0.75	0.79	0.77	0.78
alpha	0.67	0.74	0.46	0.47
burial CB14	0.55	0.57	0.35	0.35
near-backbone-11	0.49	0.54	0.24	0.25

The bits-saved measure is the one used for training, as Q_n (the fraction of most probable letters that are correct) is not really comparable between different alphabets. Note that the backbone alphabets carry more predictable information than the burial alphabets (CB14 and near-backbone-11).