HELA: A fast and accurate tool for Helitrons annotation
Résumé
Helitrons constitute a distinct superfamily of transposable elements that are supposed to employ a rolling-circle transposition mechanism. Helitron2 and Helentron are newly defined variants of Helitrons recognizable by their coding potential and terminal sequences. These subgroups are prevalent across fungi, animals, and plants genomes in which they can represent a large fraction of genome sequences. Helitrons are challenging to identify as they rarely create target site duplications upon transposition and their terminal sequences are difficult to identify. The structure-based software HelitronScanner recognize Helitrons sequences based on the TC dinucleotides at 5’ end and a hairpin structure ended with CTRR at 3’ end. This strategy suffers from extremely high false positive rate and is unable to detect Helitron2 or Helentron sequences that doesn’t necessarily start with TC dinucleotides or end with CTRR. RepeatModeler2, a commonly used software for detecting transposon-derived sequences, is unable to detect full length Helitrons if their copy number stay low, or if the elements are ancient. Here, we present HELA, a software specifically developed for the automated detection and classification of Helitron, Helentron and Helitron2 sequences within a genome. We tested HELA on a variety of genomes and checked manually the results obtained on three complex genomes: Xenopus tropicalis, Xenopus laevis and Oryza sativa. We benchmarked HELA results against a manually curated Helitrons library of Fusarium oxysporum, and compared its performance including sensitivity, precision, recall and FDR with other software including HelitronScanner, EAHelitron, and RepeatModeler2. The obtained results clearly demonstrate that HELA surpasses other software solutions and establishes itself as a valuable tool for the accurate annotation of this transposon superfamily.