Table 3.
Sensitivity, specificity and F1 score for the classification of k-mers into trusted and untrusted subsets
10× cov | 25× cov | 50× cov | |||||||||
---|---|---|---|---|---|---|---|---|---|---|---|
sens | spec | F1 | sens | spec | F1 | sens | spec | F1 | |||
S. enterica | |||||||||||
Quake | 0 | 100 | 0 | 93.77 | 100 | 96.78 | 99.31 | 99.97 | 99.65 | ||
BayesHammer | 98.22 | 99.42 | 98.81 | 99.22 | 99.98 | 99.60 | 99.42 | 99.98 | 99.71 | ||
Detox | 98.46 | 99.21 | 98.82 | 99.51 | 99.95 | 99.74 | 99.75 | 99.96 | 99.87 | ||
P. aeruginosa | |||||||||||
Quake | 4.64 | 100 | 8.88 | 80.58 | 100 | 89.24 | 98.08 | 100 | 99.03 | ||
BayesHammer | 99.07 | 99.64 | 99.11 | 99.25 | 99.99 | 99.62 | 99.23 | 100 | 99.61 | ||
Detox | 99.23 | 99.69 | 99.25 | 99.80 | 99.97 | 99.88 | 99.90 | 99.97 | 99.94 | ||
C. elegans | |||||||||||
Quake | n/a | n/a | n/a | n/a | n/a | n/a | 91.68 | 99.94 | 95.65 | ||
BayesHammer | 96.64 | 98.71 | 97.40 | 96.49 | 99.88 | 98.18 | 96.34 | 99.95 | 98.13 | ||
Detox | 99.20 | 98.96 | 98.87 | 99.81 | 99.69 | 99.82 | 99.91 | 99.68 | 99.91 | ||
A. thaliana | |||||||||||
Quake | 0 | 100 | 0 | 94.20 | 99.99 | 97.01 | 99.00 | 99.96 | 99.49 | ||
BayesHammer | 96.34 | 98.52 | 96.93 | 95.88 | 99.84 | 97.84 | 95.11 | 99.96 | 97.49 | ||
Detox | 98.36 | 97.68 | 97.68 | 99.42 | 99.40 | 99.49 | 99.70 | 99.54 | 99.76 | ||
H. sapiens chr. 21 | |||||||||||
Quake | n/a | n/a | n/a | 25.06 | 99.46 | 39.88 | 48.96 | 99.86 | 65.7 | ||
BayesHammer | 84.62 | 99.05 | 90.53 | 88.9 | 99.79 | 94.01 | 90.8 | 99.84 | 95.12 | ||
Detox | 90.09 | 98.39 | 92.88 | 95.84 | 98.47 | 97.04 | 97.5 | 98.41 | 98.2 |
Quake and BayesHammer were run with default settings. Our CRF method (Detox) was run at neighbourhood size s=3. For C. elegans at 10× and 25× coverage and H. sapiens, chr. 21, Quake was unable to identify a cutoff and hence, no results are shown (n/a). The highest values are shown in boldface