Table 1.
Results of Comparisons between TE Prediction Methods That Use Reference Sequences and the Release 3 FlyBase TE Annotations
Table 2.
Results of Comparisons between TE Prediction Methods That Do Not Use Reference Sequences and the Release 3 FlyBase TE Annotations
Figure 1.
Schematic of Our TE Annotation Pipeline
The pipeline is composed of (i) known TE family detection methods such as BLRn, RM, and RMBLR; (ii) satellite detection software such as RM, TRF, and Mreps; (iii) anonymous TE detection methods such BLRa, TE-HMM, RECON, and BLRtx; and (iv) a MySQL database called REPET to manage the results and the annotations. GAME-XML files are then generated from the results stored in the database and loaded into the Apollo genome annotation tool, allowing automatic results to be manually curated to produce a reliable annotation. To facilitate manual curation, we automatically promoted RMBLR results to be the candidate annotation.
Figure 2.
Screenshot of an Apollo View for a Peri-Centromeric Region with Extreme TE Density
Curated annotations on both forward strand (top) and reverse strand (bottom) are displayed in the light blue panels. Evidence tiers are shown in the black panels: TE-HMM (yellow), RECON (light purple), BLRa (violet), BLRtx (red), BLRn (teal), RM (blue), RMBLR (light green), and Release 3.1 FlyBase annotations (peach).
Figure 3.
Categories of Possible Boundary Comparisons between Predictions and Reference Annotations
The different cases taken into account can be grouped according to one-to-one (1-to-1), one-to-many (1-to-n), many-to-one (n-to-1), or many-to-many (n-to-n) relationships.