Superior ab initio Identification, Annotation and Characterisation of TEs and Segmental Duplications from Genome Assemblies

AbstractTransposable Elements (TEs) are mobile DNA sequences that make up significant fractions of amniote genomes. However, they are difficult to detect and annotate ab initio because of their variable features, lengths and clade-specific variants. We have addressed this problem by refining and developing a Comprehensive ab initio Repeat Pipeline (CARP) to identify and cluster TEs and other repetitive sequences in genome assemblies. The pipeline begins with a pairwise alignment using krishna, a custom aligner. Single linkage clustering is then carried out to produce families of repetitive elements. Consensus sequences are then filtered for protein coding genes and then annotated using Repbase and a custom library of retrovirus and reverse transcriptase sequences. This process yields three types of family: fully annotated, partially annotated and unannotated. Fully annotated families reflect recently diverged/young known TEs present in Repbase. The remaining two types of families contain a mixture of novel TEs and segmental duplications. These can be resolved by aligning these consensus sequences back to the genome to assess copy number vs. length distribution. Our pipeline has three significant advantages compared to other methods for ab initio repeat identification: 1) we generate not only consensus sequences, but keep the genomic intervals for the original aligned sequences, allowing straightforward analysis of evolutionary dynamics, 2) consensus sequences represent low-divergence, recently/currently active TE families, 3) segmental duplications are annotated as a useful by-product. We have compared our ab initio repeat annotations for 7 genome assemblies (1 unpublished) to other methods and demonstrate that CARP compares favourably with RepeatModeler, the most widely used repeat annotation package.Author summaryTransposable elements (TEs) are interspersed repetitive DNA sequences, also known as ‘jumping genes’, because of their ability to replicate in to new genomic locations. TEs account for a significant proportion of all eukaryotic genomes. Previous studies have found that TE insertions have contributed to new genes, coding sequences and regulatory regions. They also play an important role in genome evolution. Therefore, we developed a novel, ab initio approach for identifying and annotating repetitive elements. The idea is simple: define a “repeat” as any sequence that occurs at least twice in the genome. Our ab initio method is able to identify species-specific TEs with high sensitivity and accuracy including both TEs and segmental duplications. Because of the high degree of sequence identity used in our method, the TEs we find are less diverged and may still be active. We also retain all the information that links identified repeat consensus sequences to their genome intervals, permiting direct evolutionary analysis of the TE families we identify.