Skip to main content

Table 1 Overview of detection performance for several methods.

From: Word correlation matrices for protein sequence analysis and remote homology detection

Method avg. ROC avg. ROC50 avg. mRFP avg. # SV
WCM 1 0.8705 0.3153 0.1065 1798
WCM 2 0.8926 0.3814 0.0833 1673
WCM 3 0.8964 0.4040 0.0813 1628
WCM 4 0.9013 0.4257 0.0801 1604
WCM 5 0.9032 0.4413 0.0795 1591
WCM 6 0.9044 0.4473 0.0778 1591
WCM 7 0.9036 0.4454 0.0785 1600
WCM 8 0.9024 0.4470 0.0801 1607
WCM 9 0.9018 0.4516 0.0815 1614
WCM 10 0.9012 0.4528 0.0830 1620
LA-eig 0.9348 0.6614 0.0489 2640
ODH Monomer 0.9135 0.4554 0.0729 1601
SVM pairwise 0.9008 0.3986 0.0810 2355
Mismatch (5,1) 0.8852 0.3815 0.0949 2943
Spectrum (3) 0.8239 0.2939 0.1535 2350
Spectrum {1,2} 0.8919 0.3913 0.0798 1560
Spectrum {1,2,3} 0.8957 0.4094 0.0766 1711
Spectrum {1,2,3,4} 0.8981 0.4180 0.0769 1882
  1. Performance evaluation results of the word correlation approach (WCM K ) using several word lengths K = 1, ..10 in comparison to local alignment kernel (LA-eig) [10], Monomer Distance Histograms (ODH Monomer) [14], SVM pairwise [6], Mismatch string kernel [8], Spectrum kernel [9] and the combination of Spectrum kernels for different word lengths (see section "Results").