Table 2 Average AUC ROC scores and processing times for various PRKs

I

Pairwise Rational

PRK-Direct-Sum (K_{P R K D S-3g r a m})

0.499

15.0

[0.486, 0.512]

Kernels (PRK)

PRK-Tensor-Product (K_{P R K T-3g r a m})

0.597

16.2

[0.589, 0.605]

(3-gram)

PRK-Metric-Learning (K_{P R K M-3g r a m})

0.641

17.4

[0.633, 0.648]

PRK-Cartesian (K_{P R K C-3g r a m})

0.640

15.0

[0.632, 0.647]

II

PRKs combined

PRK-Direct-Sum+Phy (K_{P R K D S-3g r a m}+K _phy )

0.425

136.2

[0.411, 0.438]

with phylogenetic

PRK-Tensor+Phy (K_{P R K T-3g r a m} + K _phy )

0.733

135.6

[0.725, 0.741]

data (K _phy Non-

PRK-Metric+Phy (K_{P R K M-3g r a m} + K _phy )

0.761

139.2

[0.753, 0.768]

sequence kernel)

PRK-Cartesian+Phy (K_{P R K C-3g r a m} + K _phy )

0.742

132.6

[0.734, 0.749]

III

PRKs combined

PRK-D-Sum+PFAM (K_{P R K D S-3g r a m} + K _pfam )

0.493

136.2

[0.480, 0.506]

with PFAM data

PRK-Tensor+PFAM (K_{P R K T-3g r a m} + K _pfam )

0.827

136.8

[0.819, 0.834]

(K _pfam

PRK-Metric+PFAM (K_{P R K M-3g r a m} + K _pfam )

0.844

140.4

[0.837, 0.850]

Sequence kernel)

PRK-Cartesian+PFAM (K_{P R K C-3g r a m} + K _pfam )

0.842

132.0

[0.835, 0.849]