Skip to main content

Table 2 Comparison against different methods on SCOP 1.67 fold benchmark

From: A discriminative method for protein remote homology detection and fold recognition combining Top-n-grams and latent semantic analysis

Average ROC and ROC50 scores

Methods

ROC

ROC50

Source

SVM-Top-n-gram

   

n = 1

0.7778

0.649

 

n = 2

0.8130

0.642

 

n = 3

0.7960

0.628

 

SVM-Top-n-gram-combine

0.8180

0.677

 

SVM-Bprofile(Ph = 0.11)

0.8042

0.644

 

SVM-Top-n-gram-combine-LSA

0.8535

0.694

 

SVM-Bprofile-LSA(Ph = 0.11)

0.8233

0.658

 

PSI-BLAST

0.5010

0.010

[24]

SVM-Pairwise

0.7240

0.359

[24]

SVM-LA

0.8340

0.504

[24]

Gpkernel

0.8440

0.514

[24]

Mismatch

0.8140

0.467

[24]

eMOTIF

0.6980

0.308

[24]

  1. SVM-Bprofile and SVM-Top-n-gram refer to the methods based on the two building blocks: binary profiles and Top-n-grams respectively. The methods with combine suffix refer to the methods combining Top-1-grams and Top-2-grams. The methods with LSA suffix refer to the corresponding methods after latent semantic analysis. The results of the methods based on binary profiles were obtained by running the implementation of our previous work.