Manual

LTR_retriever User Manual

Shujun Ou and Ning Jiang

oushujun@msu.edu & jiangn@msu.edu

Department of Horticulture, Michigan State University, East Lansing, MI, 48824, USA

In any research documents using LTR_retriever please cite the following paper:

Shujun Ou and Ning Jiang (2017) LTR_retriever: a highly accurate and sensitive program for

Questions and Issues Please See: https://github.com/oushujun/LTR_retriever

1 Introduction

LTR_retriever is a command line program (in Perl) for accurate identification of LTR

retrotransposons (LTR-RTs) from outputs of LTRharvest (1), LTR_FINDER (2), and MGEScan-

LTR (3, 4) and generation of a non-redundant LTR-RT library for genome annotations.

accuracy and specificity needs, with several new functions to facilitate genome annotation and other

downstream studies.

LTR_retriever aims to identify high-quality LTR-RT exemplars (Figure 1A) that are intact

and non-redundant from a variety of LTR-RT candidates. To retain sensitivity, sequences of nested

LTRs and truncated LTRs (Figure 1CD) that are not represented by intact LTR-RTs will also be

included in the exemplar. This package excludes the vast majority of the non-LTR false positives. The

most common false positives were introduced by two adjacent non-LTR repeats which are found as

SINEs, LINEs, DNA TEs, or solo-LTRs that are derived from different elements (Figure 3). In

addition, LTR_retriever excludes non-LTR open reading frames derived from LINEs, DNA TEs,

or plant coding sequences to reduce misannotations of non-LTR coding sequences as LTR elements.

- A command line Perl program;

- Supports multi-threading;

- Identifies intact LTR-RTs with accurate boundaries;

- Identifies rare LTR-RTs with non-canonical (non-'TGCA') motifs;

- Supports multiple inputs: LTRharvest, LTR_FINDER, and/or MGEScan_LTR;

- Sequence input: FASTA format (contigs, scaffolds, genomes, corrected PacBio reads, and

- Output: a non-redundant LTR-RT library (FASTA), GFF3 for all intact LTR-RTs, whole-

our manual annotation, and also found in other studies (e.g., Tos17 (6) ; AtRE1 (7); and TARE1 (8)).

Flanking the terminal motifs is the target site duplication (TSD), which is generated by staggered cuts

from integrase activity (Figure 2) during LTR-RT insertion. TSDs are typically 5 bp in plants but

could vary between 3-6 bp, and the 5' and 3' TSD should be identical because of the mechanism of

their formation (Figure 2). The recently inserted LTR-RT has a highly similar LTR region that is

recognizable by sequence alignment, which is the primary searching scheme for LTR search programs

(1, 2, 4, 9). However, if two highly similar repetitive elements other than LTR (e.g., DNA, LINE,

SINE, solo-LTR, tandem repeat, etc.) are located close to each other (Figure 3), searching tools may

falsely choose them and report them as LTR-RT candidates. These are the most frequent false

positives that occur in de novo searches for LTR-RTs. Given that LTR-RTs are following the "copy-

duplications (TSD); red triangles: LTR motifs; blue boxes: long terminal repeat (LTR);

of truncated LTR-RTs. Drawing is not on scale.

two closely positioned SINE/LINE/DNA elements/solo-LTRs/tandem repeats. Light blue boxes:

but commonly has extended sequence identity on one or both termini (orange and brown

boxes).

3 Workflow of LTR_retriever

In LTR_retriever, there are eight modules developed to screen and filter out false positives

analysis and runs in UNIX-like systems. These programs include:

To run LTR_retriever, you need to provide the paths to the following dependent programs:

If the above programs are all accessible through ENV in the UNIX-like system (i.e. paths exported

For example:

Modify the following lines

Save changes to paths and exit. The installation is done.

LTR_retriever will modify sequence names that are longer than 20 characters to fit the naming

space of RepeatMasker. Thus, FASTA sequence names are recommended to be less than 20

characters without spaces and special punctuation marks other than dots (.) and underscores

(_). For long sequencing reads (e.g., PacBio), self-corrected reads are needed unless the sequence

error rate is lower than 10%.

5.2 LTR-RT candidates

LTR_retriever takes multiple LTR-RT candidate inputs including the standard output of

LTRharvest, the standard output of LTR_FINDER, and the candidate output of MGEScan-LTR.

Users need to obtain the input file(s) from the aforementioned programs before running

LTR_retriever. Either a single input source or a combination of multiple inputs is acceptable. The

following command lines provide examples and suggestions to obtain inputs.

Input from LTRharvest (a program of GenomeTools (10)): (“\” indicates this line and next

line belong to the same command line. Please delete “\” if you run into errors.)

To recover non-canonical LTR-RTs, you may use the –nonTGCA option to provide extra

7.2 WARNINGS

LTR_retriever will alter sequence names longer than 20 characters to fit the naming requirement

of RepeatMasker. Please backup your original genome file before using LTR_retriever.

LTR_retriever can take multiple sources as inputs for one single run, but running multiple

instances of LTR_retriever in the same folder at the same time may cause errors.

7.3 An overview of all parameters

Table 1. All parameters for LTR_retriever.