LTRharvest, an efficient and flexible software for de novo detection of LTR retrotransposons.

Ellinghaus D, Kurtz S, Willhoeft U. · 2008

Read LTRharvest, an efficient and flexible software for de novo detection of LTR retrotransposons. by Ellinghaus D, Kurtz S, Willhoeft U. (2008) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc2253517.txt.

Introduction

Ellinghaus, Kurtz and Willhoeft released LTRharvest in 2008, software for finding LTR retrotransposons in genome sequences. These are jumping genetic elements bordered by long terminal repeats, which makes them suited to computer detection. High-quality annotations of such elements lag behind genome sequencing. Their tool searches de novo, from scratch, using known element features such as repeat length, spacing and sequence motifs, without needing a reference library.

Against gold-standard annotations for yeast and fruit fly, the software reached sensitivity up to 90 and 97 percent respectively. Specificity was 100 percent for yeast and 72 percent for fly, matching or beating earlier tools. Its main advantages are speed on very large data sets and flexibility in using known features. It processed the largest human chromosome in about eight minutes on a standard Linux computer. The software is open source.

Catalog record

CREATOREllinghaus D, Kurtz S, Willhoeft U.
YEAR2008
COLLECTIONPAPERS
FORMATTXT
FILE SIZE52 KB
LICENSECC BY 4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/articles/PMC2253517

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

LTRharvest, an efficient and flexible software for de novo detection of LTR retrotransposons By David Ellinghaus, Stefan Kurtz, Ute Willhoeft Published in: BMC Bioinformatics DOI: 10.1186/1471-2105-9-18 ABSTRACT Background Transposable elements are abundant in eukaryotic genomes and it is believed that they have a significant impact on the evolution of gene and chromosome structure. While there are several completed eukaryotic genome projects, there are only few high quality genome wide annotations of transposable elements. Therefore, there is a considerable demand for computational identification of transposable elements. LTR retrotransposons, an important subclass of transposable elements, are well suited for computational identification, as they contain long terminal repeats (LTRs). Results We have developed a software tool LTRharvest for the de novo detection of full length LTR retrotransposons in large sequence sets. LTRharvest efficiently delivers high quality annotations based on known LTR transposon features like length, distance, and sequence motifs. A quality validation of LTRharvest against a gold standard annotation for Saccharomyces cerevisae and Drosophila melanogaster shows a sensitivity of up to 90% and 97% and specificity of 100% and 72%, respectively. This is comparable or slightly better than annotations for previous software tools. The main advantage of LTRharvest over previous tools is (a) its ability to efficiently handle large datasets from finished or unfinished genome projects, (b) its flexibility in incorporating known sequence features into the prediction, and (c) its availability as an open source software. Conclusion LTRharvest is an efficient software tool delivering high quality annotation of LTR retrotransposons. It can, for example, process the largest human chromosome in approx. 8 minutes on a Linux PC with 4 GB of memory. Its flexibility and small space and run-time requirements makes LTRharvest a very competitive candidate for future LTR retrotransposon annotation projects. Moreover, the structured design and implementation and the availability as open source provides an excellent base for incorporating novel concepts to further improve prediction of LTR retrotransposons. BACKGROUND For decades it has been known that parts of eukaryote genomes are repetitive. The major fraction of repeats are transposable elements, which are spread throughout the genomes in an interspersed fashion and make up approx. 50% of the human genome [1] or even higher percentages in plant species. Transposable elements are classified in three groups according to their mode of mobilisation (transposition): LTR retrotransposons (retrovirus like elements), non-LTR retrotransposons, and DNA transposons. They are well described by their sequence features as analysed with molecular genetics methods and by sequence comparison. However, the molecular mechanisms of transposition are in most cases not fully understood and the elucidation of their functions is still a matter of discussion. In addition, there are more questions than answers regarding the origin of transposable elements and their role in evolution, especially their contribution to modification of genomes. The genome wide identification of transposable elements is not only an essential step in the annotation of genomes but also offers an opportunity to obtain in depth knowledge on features of transposable element families. Repeat detection is a well studied problem in bioinformatics. It is mostly based on sequence analysis. A multitude of computational tools have been developed for an automated annotation of repeat families in sequenced genomes. Probably the best known program is RepeatMasker [2]. It screens DNA query sequences for interspersed repeats and low complexity DNA sequences. Precompiled sequence libraries and special scoring matrices are used to detect similar copies in the query sequence. Therefore, RepeatMasker is the first choice for repeat annotation in genomes, in which repeat families have already been well characterized, although annotating vertebrate genomes takes days of calculation time on a single computer. Transposable element families are reported to be lineage-specific, e.g. half of all human repeats arose after the divergence of mouse and human and most repeats in the mouse genome are not found in the human genome [3]. RepeatMasker's repeat libraries contain lots of repeat families from model organisms, while repeat libraries for non-model organisms exist with only limited curation [4]. In addition, transposable elements within a family might be highly divergent depending on the time of activity of the source repeat and in these cases identification by sequence comparison methods is not always successful. This holds especially when the search is performed over species borders. Therefore, automated de novo methods for repeat detection are desirable. Once for conducting a fast repeat detection, and additionally to speed up the cumbersome process of generating repeat libraries. In most cases de novo methods for finding repeats start with a self-comparison to detect sequence similarities, followed by clustering methods to group related sequences into families. A number of widely used programs are already available for this task, e.g. REPuter [5] (or the improved and more general software tool Vmatch [6]), RECON [7], RAP [8] and PILER [9]. Besides their repetitive nature, most classes of transposable elements are characterized by more specific constraints. These can be, for example, distance and length constraints for repetitive sequences within the repeat element or some motif, that is typical for this repeat element. Thus, repeats identified by general repeat detection tools additionally have to be screened in order to find candidates satisfying criteria of their specific class of transposable element. To achieve high quality predictions it is necessary to build software for the individual classes of transposable elements. LTR retrotransposons make up a large fraction of the interspersed repeats and are well classified by several structural attributes [10]. The long terminal repeats (LTRs) are the hallmark of canonical LTR retrotransposons and make them an ideal target for de novo prediction. Full length or nearly full length LTR retrotransposons bear the following features that might be suitable for de novo prediction: LTRs appear in a certain size range and distance between each other. In addition, LTRs are flanked by a short target site duplication (TSD). In some species, e.g. in yeast, the LTRs contain a conserved dinucleotide motif at their 5' and 3' end. The internal region contains genes important for retrotransposition and some conserved sequence motifs. However, depending on the age of the element, the LTR sequences, the open reading frames, and motifs are degenerated through mutations. There are already some software programs specifically designed for the de novo LTR retrotransposon detection problem. LTR_STRUC [11] is the best known of these programs. LTR_STRUC has been applied to the genomes of the fruitfly Drosophila melanogaster [12], Oryza sativa (rice) [13], Mus musculus (mouse) [14] and recently Pan troglodytes (chimpanzee) [15]. The program LTR_par [16] follows a similar approach but uses a faster algorithm than LTR_STRUC. The de novo prediction of LTRs is also the first step in the recently developed software [17] (abbreviated LTR_Rho in this publication), and by the program LTR_FINDER [18]. Both programs consider further features of LTR retrotransposons in post processing steps to enhance the quality of the predictions. The aim of this work is the development of a software tool, that efficiently works on large genomes and in addition is flexible in parameterization in order to be used for various species. Our program called LTRharvest implements the same LTR model as LTR_par, but uses a different composition of algorithms and features: 1) It allows for fast computation of large data sets, e.g. vertebrate genomes that are in the order

Read the complete text