Skip to main content
. 2011 Jun 14;27(13):i61–i68. doi: 10.1093/bioinformatics/btr249

Table 9.

AUCs of ROC curves using best-of-six gold standard

Methods SAS=2 SAS=3.5 SAS=5 Average Rank Speed
SSM using SAS score 0.94 0.9 0.89 0.91 1 13
Structural using SAS score 0.9 0.81 0.84 0.85 2 39
Combined model 0.92 0.82 0.75 0.83 3 Fast
Structural using native score 0.87 0.77 0.83 0.823 4 39
Multiview model I (400,11) 0.91 0.8 0.76 0.823 4 Fast
CE using native score 0.9 0.79 0.74 0.81 6 54
Multiview model II (400,11) 0.9 0.78 0.73 0.803 7 Fast
FragBag Cos distance (400,11) 0.89 0.77 0.75 0.803 7 Fast
Multiview model III (400,11) 0.89 0.77 0.7 0.787 9 Fast
CE using SAS score 0.84 0.72 0.75 0.77 10 54
FragBag histogram intersection (600,11) 0.87 0.73 0.7 0.767 11 Fast
SGM 0.86 0.71 0.68 0.75 12 Fast
FragBag Euclidean distance (40,6) 0.86 0.71 0.64 0.737 13 Fast
Zotenko et al. (18) 0.78 0.64 0.66 0.693 14 Fast
Sequence matching by BLAST e-value 0.76 0.57 0.5 0.61 15 Fast
PRIDE 0.72 0.54 0.51 0.59 16 Fast

The proposed approaches are shown in bold. Multiview model I is a combination of TF and LDA, multiview model II is a combination of TF-IDF and LDA and multiview model III is a combination of Boolean vector and LDA. The speed is given as average CPU minutes per query. If the processing time (after preprocessing of protein structure) for a query is < 0.1 s, then it is mentioned as fast.