Skip to main content

Table 1 Average ROC scores for motif-based kernels on SCOP 1.53 superfamily and fold benchmarks

From: Motif kernel generated by genetic programming improves remote homology and fold detection

Method

SF

Fold

Source

GPkernel

0.899

0.825

 

Mismatch

0.872

-

[17,18]

SVM-Pattern [38]

0.835

-

[38]

SVM-Pattern-LSA [38]

0.879

-

[38]

LA-eig (β = 0.2)

0 923

0 847

[18]

LA-eig (β = 0.5)

0.925

0.771

[18]

SW-PSSM(3.0,0.750,1.50)

0.982

0.933

[18]

BV-PSSM(4)

0.963

0.941

[18]

  1. SF and Fold are the average ROC-scores on the SCOP 1.53 superfamily and fold benchmark sets [12,18]; dash (-) represents an unreported value; Source is the sources of results other than ours. For reference, the table's lower part shows the best average ROC-scores reported by Rangwala and Karypis [18]. These methods do not use discrete sequence motifs.