Uncover Hidden DNA Patterns of Helitrons in Plant Genome

Helitrons are unique from other types of transposable elements because they have the capability to capture gene fragments and carry them throughout the genome.  However they are difficult to identify because they lack the classic characteristics found in other types of transposable elements.  Here we describe HelitronScanner, a generalized tool for their identification based on a motif-extracting algorithm proposed initially in a study of natural languages. We applied a two-layered Local Combinational Variable (LCV) approach for generalized Helitron identification.  LCV represents any combination pattern of either nucleotides or the associations of these nucleotide patterns. The first layer extracts location-nonspecific LCVs (n-LCV) in a known Helitron set and then creates a distribution matrix of these n-LCVs matching against Helitrons.  The second layer draws location-specific LCVs (s-LCV) from the distribution matrix.  This two-layered procedure is applied to new sequences comparably. In HelitronScanner, n-LCVs represent sequence patterns shared by Helitrons and s-LCVs constitute the associations of these patterns within Helitrons. HelitronScanner overcomes the divergence of Helitron termini among species by using conserved nucleotides at potentially variable locations.  We ran HelitronScanner against a wide range of plant genome sequences from Phytozome version 9.0 and identified a total of 107,367 Helitrons.  Many new Helitrons were identified in model species, such as maize, rice, and Arabidopsis, and in a variety of organisms where Helitrons had not been reported previously, leading to a major upward reassessment of their abundance in plant genomes. Helitron abundance varied greatly among sequenced genomes.