next_inactive up previous
Up: LTR_FINDER Home Page

version 1.0.2

Zhao XU and Hao WANG


Structure of LTR retrotransposons

Figure 1: basic structure of a full-length LTR retrotransposon
The typical structure of a full-length LTR retrotransposon is shown in Figure 1:
  1. LTR Region: 5'LTR and 3'LTR are two similar regions. They are identical while the element inserts into the host genome, and once inserted, they begin to evolve independently. Mutations and indels thus are often found. A typical LTR retrotransposon has a structure called TG..CA box, with TG at the 5' extremity of 5'LTR and CA at the 3' extremity of 3'LTR.
  2. TSR Region: TSR(target site repeat) is a 4~6bp short direct repeat string flanking the 5' and 3' extremities of an element. It is the sign of insertion of transposable elements.
  3. PBS: Near 3' end of the 5'LTR, there is a ~18bp sequence complemented to the 3' tail of some tRNA. The site is very important because tRNA binding process is the first step of initiating reverse transcription.
  4. PPT: Polypurine tract is a short rich purine segment, about 11~15 bp in length. Like PBS, this region is important for reverse transcription.
  5. Protein domains: In a typical virus genome there are three polygenes: gag, pol and env. Among them, pol is most conserved. Inside pol there are three important domains: IN(integrase), RT(reverse transcriptase) and RH(RNase H), which are enzymes for reverse transcription and insertion. RT and IN are regarded essential for autonomous LTR elements to fulfill their function.
  6. These signals may become blur or even undetectable for evolutionary events.


The Program first constructs all exact match pairs by a suffix-array based algorithm and extends them to long highly similar pairs. Then Smith-Waterman algorithm is used to adjust the ends of LTR pair candidates to get alignment boundaries. These boundaries are subject to re-adjustment using supporting information of TG..CA box and TSRs and reliable LTRs are selected. Next, LTR_FINDER tries to identify PBS, PPT and RT inside LTR pairs by build-in aligning and counting modules. RT identification includes a dynamic programming to process frame shift. For other protein domains, LTR_FINDER calls ps_scan (from PROSITE, to locate cores of important enzymes if they occur. Then possible ORFs are constructed based on that. At last, the program reports possible LTR retrotransposon models in different confidence levels according to how many signals and domains they hit.

Input data


LTR_FINDER accepts only FASTA format sequences, and only the first ungapped string(identifier) in the description line is recorded to identify the input sequence, other options in the description line will be ignored. Here is an example of input:
>CHR1 19971009 Chromosome I Sequence
... ... ... ...
>CHR2 19970727 Chromosome II Sequence
... ... ... ...

Size limit

Users could submit sequences large to 50,000,000 bytes. The timeout limit for uploading sequence is 60 minutes. For users who want to scan very large size sequences, executive binary code will be available on request.

Output format

LTR_FINDER offers three types of output: Full output, Summary output and Figure output.

Full output & Summary output

An example of Full output format is presented as follows:
>Sequence: CHR2 Len:813138
[1] CHR2 Len:813138
Location : 29632 - 35590 Len: 5959 Strand:+
Score    : 9 [LTR match score:1]
Status   : 11111111100
5'-LTR   : 29632 - 29963 Len: 332
3'-LTR   : 35259 - 35590 Len: 332
5'-TG    : TG , TG
3'-CA    : CA , CA
TSR      : 29627 - 29631 , 35591 - 35595 [ATAAT]
Sharpness: 0.479,0.52
Strand + :
PBS   : [17/22] 30031 - 30052 (ThrAGT)
PPT   : [11/15] 35215 - 35229
Domain: 31889 - 32416 [possible ORF:31193-35236, (IN (core))]
Domain: 33779 - 34387 [possible ORF:31193-35236, (RT)]

Details of exact match pairs:
35259-35472[214] (1) 35474-35590[117]
29632-29845[214] (1) 29847-29963[117]

Details of the LTR alignment(5'-end):
||  ||  |    ||| |||   |    |||*|||||||||||||||||||||||||||||

Details of the LTR alignment(3'-end):
|||||||||||||||||||||||||||||* |  |||    || ||   || | ||

Details of the PBS alignment(+):
tRNA type: ThrAGT
|||||||||    ||| |||||

Details of PPT(+):

While most of the output is straight forward, there are some fields need further explanation.

  1. Score
    Score is an integer varying from 0 to 11. It measures if signals (TSR, TG..CA box, PBS, PPT, IN(core), IN(c-term), RT, RH) occur. Because TG..CA box consists of four parts: TG at 5' end of 5'LTR, CA at 3' end of 5'LTR, TG at 5' end of 3'LTR and CA at 3' end of 3'LTR, there are 11 signals in total. The LTR match score ($match_{score}$) is the sequence similarity between 5'LTR and 3'LTR. It is a decimal between 0 and 1.
  2. Status
    Status is an 11 bits binary string, each bit indicates the status of a certain signal. From left to right, signals are: TG in 5' end of 5'LTR, CA in 3' end of 5'LTR, TG in 5' end of 3'LTR, CA in 3' end of 3'LTR, TSR, PBS, PPT, RT, IN(core), IN(c-term) and RH. If a signal occurs, corresponding position is 1 and 0 otherwise.

  3. Sharpness
    It is a decimal between 0 and 1 to evaluate the fineness of the boundary of LTR region. Higher sharpness means more accurate boundary decision. The first value is sharpness of 5' end and the second is that of 3' end. In a window of length $2W$, Sharpness of the center position is:

    \begin{displaymath}Sharpness = \frac{M_{inside}}{W}-\frac{M_{outside}}{W}\end{displaymath}

    where $M_{inside}$ and ($M_{outside}$) are the number of matched bases in left half and right half window respectively. Put the center position at the LTR boundary to get the sharpness.

  4. PBS & PPT
    For PBS, the first number in square brackets is number of matched bases and the second is total alignment length. For PPT, the first is the number of purines and length of putative PPT. Following is signal positions. String in parentheses is the tRNA type and anti-codon(See this web page for detail: The minus sign before tRNA type stands for reverse strand(not showed in this example).
  5. Details of exact match pairs
    This section shows the exact math pairs used to construct the LTR alignment. Number in square brackets is the pair length and number in parentheses is the distance between neighboring exact match pairs.
  6. Details of the LTR alignment
    This section shows the alignment details around 5' and 3' boundaries of LTR regions. Single asterisk in `|' line points out putative boundary after the second run boundary decision (see Strategy Section). Other 4~6 continuous asterisks show the positions of putative TSR.
  7. Details of PBS & PPT
    Numbers indicates the 5' ends of signals.

Summary output

Summary output is extracted from Full-output by omitting some detailed information. Here is an example output:
>Sequence: CHR2 Len:813138
[1] CHR2 Len:813138
Location : 29632 - 35590 Len: 5959 Strand:+
Score    : 9 [LTR match score:1]
Status   : 11111111100
5'-LTR   : 29632 - 29963 Len: 332
3'-LTR   : 35259 - 35590 Len: 332
5'-TG    : TG , TG
3'-CA    : CA , CA
TSR      : 29627 - 29631 , 35591 - 35595 [ATAAT]
Sharpness: 0.479,0.52
Strand + :
PBS   : [17/22] 30031 - 30052 (ThrAGT)
PPT   : [11/15] 35215 - 35229
Domain: 31889 - 32416 [possible ORF:31193-35236, (IN (core))]
Domain: 33779 - 34387 [possible ORF:31193-35236, (RT)]

Figure output

If user select output with figure, LTR_FINDER will produce a PNG file to show the relative position of each LTR retrotransposons. The figure was plotted by both normal axis and logarithmic axis. Elements drawn on silver background was plotted by their real size, that means, 1 pixel stand for 1 base. Elements drawn on white background was plotted under logarithmic axis, so that the long distance could be resized to place on a small canvas. Blue circles denote for PBS, brown circles stand for PPT, and purple circles on the end of LTRs are TSRs.
Figure 2: Legend of output figure


LTR_FINDER has many parameters, which can be divided into two groups: parameters used in finding LTR retrotransposons (construction parameters) and parameters used in filtering out unreliable results (filter parameters). The first group includes -o, -t, -e, -m, -u, -D, -d, -L, -l, -p, -g, -J, -j, -s, -a and -r; the second group includes -S, -B, -b, -w, -O, -P and -F.

-o, Gap open penalty, (positive integer)
-t, Gap extend penalty, (positive integer)
-e, Gap end penalty, (positive integer)
-m, Match score, (positive integer)
-u, Mismatch score, (negative integer)

The five parameters control alignment algorithm. Denote gap open penalty as $P_{open}$, gap extend penalty as $P_{ext}$, gap end penalty as $P_{end}$, match score as $S_{match}$ and mismatch score as $S_{mismatch}$, then global and local alignments score are :

Score_{local} & = & S_{match}\cdot N_{match}+S_{mismatch}\cdot...
...ismatch} \\
& & - \sum P_{inner\_gap} - P_{5'gap}- P_{3'gap}\\

where $N_{match}$ is the number of match bases; $N_{mismatch}$ the number of unmatched bases; and

P_{inner\_gap} & = & P_{open}+P_{ext}\cdot(inner\_gap\_len-1)\...
...d}\cdot 5'gap\_len\\
P_{3'gap} & = & P_{end}\cdot 3'gap\_len\\

where $inner\_gap\_len$ is the length of gap minus 2 (end bases). Usually, $P_{open}$ is greater than $S_{match}$.

-D, Max distance between LTRs, (positive integer)
-d, Min distance between LTRs, (positive integer)
-L, Max LTR Length, (positive integer)
-l, Min LTR Length, (positive integer)

Distance between LTRs is:


The four parameters makes detected models meet common features of LTR retrotransposons.

-g, Extension max gap, (positive integer)
-j, Extension cutoff, (decimal between 0 and 1)
-J, Reliable extension, (decimal between 0 and 1)

The three parameters control the extension of LTRs. An example of 2 neighbouring LTR pairs is shown in Figure 3. If $s[i_b\ldots m_e]$ and $s[j_b\ldots n_e]$ are similar enough, we extend LTRs from $s[i_b\ldots i_e]$ and $s[m_b\ldots m_e]$ to $s[i_b\ldots m_e]$ and $s[j_b\ldots n_e]$.

Figure 3: $P_1\{s[i_b\ldots i_e], s[j_b\ldots j_e]\}\ and \ P_2\{s[m_b\ldots m_e], s[n_b\ldots n_e]\}$

$P_1$ and $P_2$ are pre-sorted so that $j_b \geq i_b$, $n_b \geq
m_b$ and $j_b \geq m_b$. Since $P_1$ and $P_2$ are two exact match pairs, we know

len_{pair1}= & i_e-i_b+1= & j_e-j_b+1\\
len_{pair2}= & m_e-m_b+1= & n_e-n_b+1

Obviously, the gap lengthes between them are


Introduce $Diff$, number of base differences resulting from extension, as

\begin{displaymath}Diff= \left\{ \begin{array}{ll}
Length_{inner\_mis} & gap_1>0...
...1,gap_2\}-min\{gap_1,gap_2\} & otherwise\\

where $Length_{inner\_mis}$ is the number of different(mismatches and indels) bases from global alignment of $s[i_{e}+1\ldots
m_{b}-1]$ and $s[j_{e}+1\ldots n_{b}-1]$. The similarity of merged loci is then:

\begin{displaymath}Sim = \frac{len_{pair1}+max\{gap_1,gap_2\}+len_{pair2}-Diff}{len_{pair1}+max\{gap_1,gap_2\}+len_{pair2}}\end{displaymath}

When LTR_FINDER decides whether two neighboring pairs should be merged, it first calculates $Diff$, make sure that it does not exceed the value of extension max gap, then calculates $Sim$. If $Sim < extension\ cutoff$, pair extension will stop here, $P_1\{s[i_b\ldots i_e], s[j_b\ldots j_e]\}$ will be reported as a candidate for LTR element; If $Sim > reliable\ extension$, new pair $P_2$ and inter-pair regions will be linked to the previous one $P_1$ to construct a longer new pair $P\{s[i_b\ldots m_e], s[j_b\ldots n_e]\}$, and LTR_FINDER continues to find next neighboring pairs; if $extension\ cutoff < Sim < reliable\ extension$, it means we are not sure whether continue to extend or stop. So LTR_FINDER first report a LTR element candidate $P_1\{s[i_b\ldots i_e], s[j_b\ldots j_e]\}$ while at the same time, the extension process will continue.

-p, Length of exact match pairs, (positive integer)

Running time of LTR_FINDER is very sensitive to this parameter. The program only selects pairs that are longer than this value to do further processes from all exact match pairs detected. If a very small value is given, LTR_FINDER will spend much time on randomly matched short sequences. We use P-value to estimate proper value of -p, that is, the probability of exact match of length longer than $L$ occurs if 2 sequences are drown randomly. In Waterman 1989(, ), under independent letter model, using asymptotic extreme value distribution, P-value were worked out analytically. Now if one assign the P-value, length can be deduced. From experiments, 20 is appropriate in most situation, and we suggest not using of value less than 15.

-r, Min match length for PBS detection, (positive integer)

When aligning tRNA 3' tail 18nt string to the inter-LTR sequence, if LTR_FINDER finds an alignment that match length exceeds this threshold, it will report this region as a putative primer binding site.

-s, Predict PBS by using which tRNA database, (filename)

To predict primer binding site, we need tRNA sequences, especially the 3' end 18nt of each sequences. LTR_FINDER can load tRNA of different species. A good tRNA set can be found at Genomic tRNA Database ( Our database was download from Genomic tRNA Database on 2007-07-17. Here is an example of required format:

>Athal-chr4.trna25-AlaAGC (13454563-13454635)
LTR_FINDER uses the string after the last minus sign in sequence identifier field as the tRNA type name. In this example, `AlaAGC'.

-a, Use ps_scan to predict IN(core), IN(c-term) and RH, dirname

This parameter is a directory name. LTR_FINDER can predict protein domains by calling ps_scan, which can be obtained from ExPASy-PROSITE ( User should place data file `prosite.dat' and ps_scan in this directory. If this parameter is enabled, LTR_FINDER will call them and report these protein domains if they are detected.

-S, Output score threshold, (integer)

This is the threshold for LTR retrotransposon score. Models that have higher scores are output.

-B, Sharpness higher threshold, (decimal between 0 and 1)
-b, Sharpness lower threshold, (decimal between 0 and 1)

LTR_FINDER calculates sharpness for both 5' and 3' extremities of putative elements. Both of them must be greater than the lower threshold and one greater than the higher threshold.

-w, Output format, (0,1 or 2)

This parameter controls the output format:

Value Format
0 Full output
1 Summary output
2 Table output(not available on web)

-O, Length of alignment details, (positive integer)

LTR_FINDER can output alignment details of LTR boundaries (left to 120bp and right to 80bp relative to boundaries). Users are allowed to assign the output length by setting this parameter. The whole 200 bp alignment is output when it $\geq 120$.

-P, Sequence ID pattern, (POSIX regular expression)

This parameter is ID of one sequence in a multi-fasta file. When enabled, only sequences with this ID will be processed.

-F, Signals status control, (01 string)

The parameter controls output by sequence tag status. It is a binary string of 11 bits. From left to right, bits denote the status of following signals: TG in 5' end of 5'LTR, CA in 3' end of 5'LTR, TG in 5' end of 3'LTR, CA in 3' end of 3'LTR, TSR, PBS, PPT, RT, IN(core), IN(c-term) and RH. 1 means reported models should containing the signal and 0 ignoring searching it. when used, the program will only report models whose status of sequence tags match it.

-C, Auto mask highly repeated regions

With this parameter, LTR_FINDER will try to mask highly repeated regions defined by: the same exact match pair repeat 14 more times within 3000bp. By using this parameter, LTR_FINDER can perform much more quickly on sequences which have centriole or telomere region.


We thank Heng Li for his linear-space pairwise alignment library and Xiaoli Shi for providing rice tRNA sequences. The authors are also grateful to colleagues who helped us testing the web server.


Flavell, R. B. (1986).
Repetitive dna and chromosome evolution in plants.
Philos Trans R Soc Lond B Biol Sci, 312(1154), 227-242.

Gao, L., McCarthy, E. M., Ganko, E. W., and McDonald, J. F. (2004).
Evolutionary history of oryza sativa ltr retrotransposons: a preliminary survey of the rice genome sequences.
BMC Genomics, 5(1), 18.

Gao, X., Havecker, E. R., Baranov, P. V., Atkins, J. F., and Voytas, D. F. (2003).
Translational recoding signals between gag and pol in diverse ltr retrotransposons.
RNA, 9(12), 1422-1430.

Jordan, I. K. and McDonald, J. F. (1998).
Evidence for the role of recombination in the regulatory evolution of saccharomyces cerevisiae ty elements.
J Mol Evol, 47(1), 14-20.

Kalyanaraman, A. and Aluru, S. (2006).
Efficient algorithms and software for detection of full-length ltr retrotransposons.
J Bioinform Comput Biol, 4(2), 197-216.

Kim, J. M., Vanguri, S., Boeke, J. D., Gabriel, A., and Voytas, D. F. (1998).
Transposable elements and genome organization: a comprehensive survey of retrotransposons revealed by the complete saccharomyces cerevisiae genome sequence.
Genome Res, 8(5), 464-478.

Ma, J., Devos, K. M., and Bennetzen, J. L. (2004).
Analyses of ltr-retrotransposon structures reveal recent and rapid genomic dna loss in rice.
Genome Res, 14(5), 860-869.

McCarthy, E. M. and McDonald, J. F. (2003).
Ltr_struc: a novel search and identification program for ltr retrotransposons.
Bioinformatics, 19(3), 362-367.

McCarthy, E. M. and McDonald, J. F. (2004).
Long terminal repeat retrotransposons of mus musculus.
Genome Biol, 5(3), R14.

McCarthy, E. M., Liu, J., Lizhi, G., and McDonald, J. F. (2002).
Long terminal repeat retrotransposons of oryza sativa.
Genome Biol, 3(10), RESEARCH0053.

McDonald, J. F. (1993).
Evolution and consequences of transposable elements.
Curr Opin Genet Dev, 3(6), 855-864.

McDonald, J. F., Matyunina, L. V., Wilson, S., Jordan, I. K., Bowen, N. J., and Miller, W. J. (1997).
Ltr retrotransposons and the evolution of eukaryotic enhancers.
Genetica, 100(1-3), 3-13.

SanMiguel, P., Tikhonov, A., Jin, Y. K., Motchoulskaia, N., Zakharov, D., Melake-Berhan, A., Springer, P. S., Edwards, K. J., Lee, M., Avramova, Z., and Bennetzen, J. L. (1996).
Nested retrotransposons in the intergenic regions of the maize genome.
Science, 274(5288), 765-768.

Xiong, Y. and Eickbush, T. H. (1990).
Origin and evolution of retroelements based upon their reverse transcriptase sequences.
EMBO J, 9(10), 3353-3362.

Yoder, J. A., Walsh, C. P., and Bestor, T. H. (1997).
Cytosine methylation and the ecology of intragenomic parasites.
Trends Genet, 13(8), 335-340.

Zhang, X. and Wessler, S. R. (2004).
Genome-wide comparative analysis of the transposable elements in the related species arabidopsis thaliana and brassica oleracea.
Proc Natl Acad Sci U S A, 101(15), 5589-5594.

About this document ...

version 1.0.2

This document was generated using the LaTeX2HTML translator Version 2008 (1.71)

Copyright © 1993, 1994, 1995, 1996, Nikos Drakos, Computer Based Learning Unit, University of Leeds.
Copyright © 1997, 1998, 1999, Ross Moore, Mathematics Department, Macquarie University, Sydney.

The command line arguments were:
latex2html -up_url -up_title 'LTR_FINDER Home Page' -transparent -antialias_text -antialias -image_type gif -local_icons -split 0 help

The translation was initiated by on 2009-04-09

next_inactive up previous
Up: LTR_FINDER Home Page