Statistics on bSiteFinder

Result comparison

Two widely adopted datasets including 210 bound and 48 bound/unbound dataset (Huang and Schroeder, 2006) were used for testing our algorithm, and the results are shown in Table 1 and Table 2. The accuracy of our algorithm is approximately 2%-10% higher than that of other algorithms for the test with either bound or unbound datasets. In addition, with size of the dataset increased, our algorithm exhibited even more advantage over others regarding accuracy (The accuracy differences between our algorithm and the second highest algorithm in the Top1 increase from 2.4% with 48 unbound dataset to 11.8% with 210 unbound dataset).

Table 1. Comparison of the top1 and top3 success rates for various methods using 210 bound structures

Method

Top1a

Top3a

bSiteFinder

94.8%

95.7%

LISEb

83%

94%

MPK2b

81%

95%

MPK1b

75%

93%

Q-SiteFinderb

70%

90%

LIGSITECSCb

75%

-

LIGSITECSb

70%

86%

PASSb

51%

80%

SURFNETb

42%

57%

 

(a The MCC scores of the Top1 and Top3 are 0.95 and 0.97 respectively with 210 bound structures. b The success rates of these methods were taken from Xie and Hwang (Xie and Hwang, 2012).)

Table 2. Comparison of the top1 and top3 success rates for various methods using 48 bound/unbound structures

Method

Bound a

Unbound b

Top1

Top3

Top1

Top3

bSiteFinder

93.8%

98.7%

85.4%

95.8%

LISEc

92%

96%

81%

92%

MPK2c

85%

96%

80%

94%

VICEc

85%

94%

83%

90%

MPK1c

83%

96%

75%

90%

DoGSitec

83%

92%

71%

92%

Fpocketc

83%

92%

69%

94%

LIGSITECSc

81%

92%

71%

85%

LIGSITECSCc

79%

-

71%

-

MSPocketc

77%

94%

75%

88%

POCASAc

77%

90%

75%

92%

Q-SiteFinderc

75%

90%

52%

75%

PocketPickerc

72%

85%

69%

85%

CASTc

67%

83%

58%

75%

PASSc

63%

81%

60%

71%

SURFNETc

54%

78%

52%

75%

(a The MCC scores of the Top1 and Top3 are 0.95 and 0.97 respectively with 48 bound structures. b The MCC scores of the Top1 and Top3 are 0.72 and 0.75 respectively with 48 unbound structures. c The success rates of these methods were taken from Xie and Hwang (Xie and Hwang, 2012).)

User distribution