Table 1.
Gene prediction accuracy of BRAKER1 and MAKER2 (both pipelines used repeat masking) as assessed by comparison with annotation of the genomes of four model organisms
|
Arabidopsis thaliana
|
Caenorhabditis elegans
|
Drosophila melanogaster
|
Schizosaccharomyces pombe
|
||||||
|---|---|---|---|---|---|---|---|---|---|
| BRAKER1 | MAKER2 | BRAKER1 | MAKER2 | BRAKER1 | MAKER2 | BRAKER1 | MAKER2 | CodingQuarry | |
| Gene sensitivity | 64.4 | 51.3 | 55.0 | 41.0 | 64.9 | 55.2 | 77.4 | 42.8 | 79.7 |
| Gene specificity | 52.0 | 52.5 | 55.2 | 30.8 | 59.4 | 46.3 | 80.5 | 68.7 | 72.6 |
| Transcript sensitivity | 55.0 | 43.5 | 43.0 | 31.3 | 46.1 | 38.5 | 77.4 | 42.8 | 79.7 |
| Transcript specificity | 50.9 | 52.5 | 53.2 | 30.8 | 57.9 | 46.3 | 76.5 | 68.7 | 72.6 |
| Exon sensitivity | 82.9 | 76.1 | 80.2 | 69.4 | 75.0 | 66.5 | 83.2 | 50.1 | 79.6 |
| Exon specificity | 79.0 | 76.1 | 85.3 | 62.3 | 81.7 | 66.9 | 83.2 | 71.4 | 81.7 |
Bold font highlights the highest accuracy reached in a given category (Sn or Sp) for a given species. In all cases, RNA-Seq was the only source of extrinsic evidence. For the fungus S.pombe, we also assessed the accuracy of gene predictions made by CodingQuarry.