Table 2 Mean and standard deviation for gene prediction performance of our method (Neural Net) and MetaGene. Performance was measured on 700 bp fragments that were randomly excised from each test genome to 5-fold coverage (ten replications per species). The harmonic mean is a measure that combines sensitivity and specificity.
From: Gene prediction in metagenomic fragments: A large scale machine learning approach
| Â | SENSITIVITY | SPECIFICITY | HARMONIC MEAN | |||
|---|---|---|---|---|---|---|
Species | Neural Net | MetaGene | Neural Net | MetaGene | Neural Net | MetaGene |
Archaeoglobus fulgidus | 87.2 ± 0.21 | 93.7 ± 0.15 | 93.4 ± 0.16 | 92.7 ± 0.16 | 90.2 ± 0.17 | 93.2 ± 0.14 |
Methanococcus jannaschii | 91.7 ± 0.17 | 95.8 ± 0.14 | 96.2 ± 0.13 | 92.7 ± 0.19 | 93.9 ± 0.10 | 94.3 ± 0.15 |
Natronomonas pharaonis | 87.9 ± 0.22 | 95.1 ± 0.09 | 93.9 ± 0.10 | 92.7 ± 0.17 | 90.8 ± 0.16 | 93.9 ± 0.12 |
Buchnera aphidicola | 90.6 ± 0.37 | 96.7 ± 0.24 | 95.3 ± 0.31 | 91.1 ± 0.29 | 92.9 ± 0.28 | 93.8 ± 0.21 |
Burkholderia pseudomallei | 87.9 ± 0.11 | 94.1 ± 0.11 | 90.1 ± 0.09 | 85.1 ± 0.13 | 89.0± 0.08 | 89.4 ± 0.10 |
Bacillus subtilis | 91.4 ± 0.16 | 89.8 ± 0.14 | 95.3 ± 0.09 | 89.3 ± 0.19 | 93.3 ± 0.10 | 89.5 ± 0.14 |
Corynebacterium jeikeium | 89.7 ± 0.24 | 91.9 ± 0.12 | 93.8 ± 0.19 | 89.2 ± 0.21 | 91.7 ± 0.19 | 90.5 ± 0.13 |
Chlorobium tepidum | 82.1 ± 0.25 | 85.7 ± 0.27 | 91.2 ± 0.17 | 88.4 ± 0.26 | 86.4 ± 0.19 | 87.0 ± 0.22 |
Escherichia coli | 91.7 ± 0.16 | 93.3 ± 0.07 | 95.3 ± 0.09 | 90.9 ± 0.10 | 93.5 ± 0.12 | 92.1 ± 0.07 |
Helicobacter pylori | 92.1 ± 0.11 | 90.2 ± 0.14 | 96.6 ± 0.15 | 89.6 ± 0.23 | 94.3 ± 0.11 | 89.9 ± 0.15 |
Pseudomonas aeruginosa | 90.4 ± 0.14 | 96.2 ± 0.07 | 92.5 ± 0.11 | 91.4 ± 0.09 | 91.4 ± 0.12 | 93.7 ± 0.07 |
Prochlorococcus marinus | 87.2 ± 0.21 | 93.7 ± 0.25 | 95.9 ± 0.14 | 90.8 ± 0.20 | 91.4 ± 0.15 | 92.2 ± 0.19 |
Wolbachia endosymbiont | 87.2 ± 0.27 | 90.6 ± 0.42 | 85.2 ± 0.44 | 71.2 ± 0.54 | 86.2 ± 0.29 | 79.7 ± 0.45 |