Table 4 |
||||||
|
Function prediction performance of motif ensembles versus single-structure motifs at significance threshold of α = 0.01. |
||||||
|
Single structure motif |
Motif ensemble (CV) |
Improvement (x-fold) |
||||
|
EC class |
||||||
|
%Sens. (#TP) |
%Spec. (#FP) |
%Sens. (#TP) |
%Spec. (#FP) |
Sens. |
Spec. |
|
|
|
||||||
|
1.1.1.1 |
52.4% (43) |
99.2% (83) |
74.3 ± 7.0% (61) |
99.4 ± 0.0% (62 ± 4) |
1.4 |
1.0 |
|
1.1.1.21 |
93.3% (83) |
99.1% (146) |
93.2 ± 4.8% (83) |
99.2 ± 0.1% (136 ± 5) |
1.0 |
1.0 |
|
1.11.1.7 |
91.6% (76) |
99.1% (131) |
92.7 ± 10.0% (77) |
99.5 ± 0.0% (78 ± 8) |
1.0 |
1.0 |
|
1.14.13.39 |
90.5% (114) |
99.3% (87) |
96.1 ± 2.7% (121) |
99.4 ± 0.0% (73 ± 7) |
1.1 |
1.0 |
|
2.5.1.18 |
25.3% (48) |
99.1% (171) |
46.3 ± 5.1% (88) |
99.2 ± 0.0% (140 ± 5) |
1.8 |
1.0 |
|
2.6.1.1 |
66.7% (70) |
99.1% (153) |
82.9 ± 5.4% (87) |
99.3 ± 0.0% (121 ± 5) |
1.2 |
1.0 |
|
2.7.4.6 |
81.7% (49) |
99.2% (137) |
88.3 ± 2.6% (52) |
99.4 ± 0.1% (113 ± 5) |
1.1 |
1.0 |
|
3.1.1.7 |
98.2% (108) |
99.2% (82) |
99.0 ± 2.0% (108) |
99.4 ± 0.0% (60 ± 2) |
1.0 |
1.0 |
|
3.1.3.1 |
84.1% (37) |
99.1% (122) |
100.0 ± 0.0% (44) |
99.3 ± 0.0% (97 ± 6) |
1.2 |
1.0 |
|
3.1.3.48 |
28.6% (71) |
99.1% (155) |
56.1 ± 3.6% (139) |
99.4 ± 0.1% (109 ± 11) |
2.0 |
1.0 |
|
3.2.1.1 |
83.5% (111) |
99.1% (149) |
88.7 ± 7.9% (117) |
99.4 ± 0.1% (102 ± 17) |
1.1 |
1.0 |
|
3.5.2.6 |
35.0% (89) |
99.2% (144) |
81.2 ± 6.3% (208) |
99.4 ± 0.0% (107 ± 9) |
2.3 |
1.0 |
|
4.2.1.1 |
87.9% (248) |
99.1% (112) |
95.3 ± 3.5% (269) |
99.6 ± 0.0% (49 ± 4) |
1.1 |
1.0 |
|
5.3.1.1 |
78.9% (75) |
99.1% (143) |
82.1 ± 10.9% (78) |
99.4 ± 0.1% (100 ± 11) |
1.0 |
1.0 |
|
5.3.1.5 |
97.3% (71) |
99.1% (118) |
98.5 ± 2.3% (71) |
99.4 ± 0.1% (92 ± 11) |
1.0 |
1.0 |
|
|
||||||
|
For each single-structure motif, a motif ensemble was constructed using FASST-MESH. Next to each % sensitivity value is the total number of true positive (TP) matches; next to each % specificity value is the total number of false positive (FP) matches. The performance of motif ensembles was assessed using 5-fold cross-validation and the sensitivity/specificity values correspond to mean ± standard deviation across the 5 folds. The x-fold improvement is calculated as: mean motif ensemble performance divided by single-structure performance. |
||||||
|
Bryant et al. BMC Bioinformatics 2010 11:242 doi:10.1186/1471-2105-11-242 |
||||||
