# 1 Introduction

The goal of packFinder was to implement a simple tool for the prediction of potential Pack-TYPE elements. packFinder uses the following prior knowledge, provided by the user, to detect transposons:

1. Terminal Inverted Repeat (TIR) Base Sequence
2. Length of Terminal Site Duplication (TSD)
3. Length of the Transposon

These features, shown in Figure 1, provide enough information to detect autonomous and pack-TYPE elements. For a transposon to be predicted by packFinder its TSD sequences must be identical to each other, its forward TIR sequence must match the base sequence provided and its reverse TIR sequence must match its reverse complement.

Transposons are therefore predicted by searching a given genome for these characteristics, and further analysis steps can reveal the nature of these elements - while the packFinder tool is sensitive for the detection of transposons, it does not discriminate between autonomous and Pack-TYPE elements.

Autonomous elements will contain a transposase gene within the terminal inverted repeats and tend to be larger than their Pack-TYPE counterparts; pack-TYPE elements instead capture sections of host genomes. Following cluster analysis, BLAST can be used to discern which predicted elements are autonomous (transposase-containing) and with are true Pack-TYPE elements.

# 2 Getting Started

Users may download packFinder and use the primary function - packSearch - to locate potential transposons in a given set of DNA sequences. In addition to R packages, the command line tool VSEARCH must be installed prior to use of clustering and alignment functions.

## 2.1 R Package Dependencies

CRAN and Bioconductor packages will be automatically installed when downloading packFinder. The package may be installed from the development branch of Bioconductor, as long as R version 4.0.0 is installed.

if (!requireNamespace("BiocManager", quietly = TRUE))
install.packages("BiocManager")

# The following initializes usage of Bioc devel
BiocManager::install(version='devel')

BiocManager::install("packFinder")

## 2.2 Command Line Dependencies

While this allows R to run the packSearch pipeline within packFinder, VSEARCH must be installed for use of clustering and alignment functions within packSearch. Detailed installation instructions are available from the README file on the VSEARCH github https://github.com/torognes/vsearch). The command line can be used to install VSEARCH on Linux and MacOS operating systems (using wget and tar) while VSEARCH can be downloaded and extracted for use on Windows systems.

For Linux and MacOS systems, correct installation of VSEARCH should allow users to use all functions within packFinder whereas for windows users, the absolute path to the VSEARCH executable file must be specified when calling packFinder clustering and alignment functions. This vignette will assume a MacOS/Linux operating system - see ?packClust for examples using windows.

# 3 Searching for Potential Transposable Elements using packFinder

## 3.1 Getting Data

Both DNA to be searched and the TIR query must be in XString format (see Biostrings). The TIR query should be coerced to a DNAString object while the DNA sequence, or set of sequences, should be a DNAStringSet. Bioconductor’s Biostrings provides methods for the conversion of various formats to DNAString objects, including from character vectors and fasta files.

# Convert a character vector to a DNAString
Biostrings::DNAString("CATG")
## 4-letter DNAString object
## seq: CATG
# Convert a list of character vectors to a DNAStringSet
Biostrings::DNAStringSet(c(
"CATG",
"GTAC"
))
## DNAStringSet object of length 2:
##     width seq
## [1]     4 CATG
## [2]     4 GTAC
# Convert a FASTA file to a DNAStringSet
system.file("extdata", "packMatches.fasta", package = "packFinder"),
format = "fasta"
)
## DNAStringSet object of length 6:
##     width seq                                               names
## [1]  1518 CACTACAAAAAAAAAGTCGAATT...AAAATGATCACTTTTTTGTAGTG Chr3 | start = 10...
## [2]   713 CACTACAAGGACAATGAAATTTA...CGTTTACCTGTTTTCTTGTAGTG Chr3 | start = 10...
## [3]  1708 CACTACAAAACAAAGTGGATTAC...TTAAAATGATCTTTTTTGTAGTG Chr3 | start = 28...
## [4]   728 CACTACAAGGACAATGAAATTTT...CTTTTACCTGTTTTCTTGTAGTG Chr3 | start = 34...
## [5]  2266 CACTACAACAAATATCCACATTC...TAATGTGGATATTTGTTGTAGTG Chr3 | start = 35...
## [6]  2463 CACTACAAGAATATTAAACTTTA...GAAAAGTCATTTTTCTTGTAGTG Chr3 | start = 37...

## 3.2 Using packSearch

In this example we search a subset of chromosome 3, from the Arabidopsis thaliana reference sequence; our query will be first 8 bases of the CACTA1 autonomous transposable element’s TIR. The CACTA transposons have TSD sequences that are 3 bases long, and the Pack-CACTA elements tend to be between 300 to 3500 bases in width. Since the first 8 bases of the Pack-CACTA TIRs tend to be conserved between elements, the default allowable mismatch of 0 was used.

data("arabidopsisThalianaRefseq")

packMatches <- packSearch(
Biostrings::DNAString("CACTACAA"),
arabidopsisThalianaRefseq,
elementLength = c(300, 3500),
tsdLength = 3
)
## Getting forward matches
## 90 forward matches identified.
## Getting reverse matches
## 97 reverse matches identified.
## Filtering matches based on TSD sequences
## Initial filtering complete. 6 elements predicted.
head(packMatches)
##   seqnames   start     end width strand TSD
## 1     Chr3  100830  102347  1518      * TGT
## 2     Chr3 1068802 1069514   713      * ATA
## 3     Chr3 2807747 2809454  1708      * GGT
## 4     Chr3 3487540 3488267   728      * ATA
## 5     Chr3 3582297 3584562  2266      * ATA
## 6     Chr3 3738747 3741209  2463      * TTG

In this subset, we identified six potential Pack-TYPE elements, however at this stage it is unclear what genetic information can be found between the inverted TIR sequences Figure 1. These elements could contain transposase genes, making them autonomous elements, or may be Pack-TYPE elements that have captured parts of the Arabidopsis thaliana host genome. Additionally, of the Pack-TYPE elements detected, some may share sequences of related chromosomal origin between their TIRs.

packSearch returns a dataframe of ranges, in the format produced by coercing a GRanges object to a dataframe: data.frame(GRanges). This format is used to transfer the transposon ranges between the functions in packFinder.

# 4 Analysing Potential Transposable Elements using packFinder

## 4.1 Clustering of Transposable Elements

In order to make downstream analysis more efficient and understand the relations between the identified elements, we can use VSEARCH to cluster predicted transposons. Here we run VSEARCH with the default parameters, meaning:

An identity of 60% between two elements is required to form a cluster The method of identity detection is the VSEARCH default

The identity and method of identity definition can be altered depending on analysis (see VSEARCH documentation).

packMatches <- packClust(
packMatches,
arabidopsisThalianaRefseq,
saveFolder = "tempOutput"
)
## Rognes T, Flouri T, Nichols B, Quince C, Mahe F (2016)
## VSEARCH: a versatile open source tool for metagenomics
## PeerJ 4:e2584 doi: 10.7717/peerj.2584 https://doi.org/10.7717/peerj.2584
##
## vsearch v2.14.1_win_x86_64, 7.9GB RAM, 4 cores
## https://github.com/torognes/vsearch
##
## vsearch v2.14.1_win_x86_64, 7.9GB RAM, 4 cores
## https://github.com/torognes/vsearch
##
## 9396 nt in 6 seqs, min 713, max 2463, avg 1566
## Counting k-mers 100%
## Clustering 100%
## Sorting clusters 100%
## Writing clusters 100%
## Clusters: 5 Size min 1, max 2, avg 1.2
## Singletons: 4, 66.7% of seqs, 80.0% of clusters
## Sorting clusters by abundance 100%
##   seqnames   start     end width strand TSD cluster
## 1     Chr3  100830  102347  1518      + TGT       3
## 2     Chr3 1068802 1069514   713      + ATA       4
## 3     Chr3 2807747 2809454  1708      + GGT       2
## 4     Chr3 3487540 3488267   728      + ATA       4
## 5     Chr3 3582297 3584562  2266      + ATA       1
## 6     Chr3 3738747 3741209  2463      + TTG       0

Of the 6 elements identified in this data subset, only two were found to be in the same cluster. When a new cluster is created, the transposon is designated as being on the forwards strand (+); elements that are subsequently assigned to this cluster are given a strand designation relative to the original element in the cluster.

For this data, as there are few clusters, all of the elements have been designated as being on the forwards strand. Adjusting the identity % required or changing the identity definition could lead to more effective clustering, or lead to false-positives.

Note, by default filterWildcards is called when clustering or aligning sequences; this prevents low quality sequences from clustering together, and by default removes sequences with a proportion of wildcards (“N”) above 5%.

## 4.2 Reading VSEARCH Output Files

Based on the results of packClust, we found that the second and fourth matches have similar sequences. Additionally, these potential elements have a similar width and so it is feasible that these are elements that have been duplicated by a transposase. To investigate the extent of the similarities, we can read the more detailed VSEARCH output files:

• USEARCH cluster format .uc - containing a summary of the clustering done by VSEARCH
• BLAST output - a BLAST compatible summary containing details of the BLAST matches between clusters
readUc(system.file(
"extdata",
"packMatches.uc",
package = "packFinder"
))
##    type cluster width identity strand  cigarAlignment query target
## 1     S       0  2463        *      *               *     6      *
## 2     S       1  2266        *      *               *     5      *
## 3     S       2  1708        *      *               *     3      *
## 4     S       3  1518        *      *               *     1      *
## 5     S       4   728        *      *               *     4      *
## 6     H       4   713     84.1      + 94MD117M16I501M     2      4
## 7     C       0     1        *      *               *     6      *
## 8     C       1     1        *      *               *     5      *
## 9     C       2     1        *      *               *     3      *
## 10    C       3     1        *      *               *     1      *
## 11    C       4     2        *      *               *     4      *
readBlast(system.file(
"extdata",
"packMatches.blast6out",
package = "packFinder"
))
##   query_id subject_id identity alignment_length mismatches gap_opens q.start
## 1        2          4     84.1              729         99         2       1
##   q.end s.start s.end evalue bit_score
## 1   713       1   728     -1         0

## 4.3 Clustering of TIR Sequences

Additionally, tirClust can provide a summary of the similarities between the TIR sequences of clustered transposons. While in this example “CACTACAA” has been used as the TIR search query, the CACTA TIR sequence is longer than 8 base pairs - although the rest of the TIR sequence may be less well conserved.

For each cluster, tirClust creates a consensus sequence for the forward and reverse TIR regions; in this case we will consider the first 25 base pairs of each TIR. Additionally, clustering of these TIRs is carried out using kmer clustering before being plotted as a dendrogram for visualisation.

consensusSeqs <- tirClust(packMatches,
arabidopsisThalianaRefseq,
tirLength = 25
)

head(consensusSeqs)
## DNAStringSet object of length 6:
##     width seq                                               names
## [1]    26 CACTACAAAAAAAAAGTCGAATTGAA                        f3
## [2]    26 CACTACAAGGACAATGAAATTTWCCT                        f4
## [3]    26 CACTACAAAACAAAGTGGATTACATC                        f2
## [4]    26 CACTACAACAAATATCCACATTCTTA                        f1
## [5]    26 CACTACAAGAATATTAAACTTTAATA                        f0
## [6]    26 CACTACAAAAAAGTGATCATTTTATA                        r3

As expected, the forward and reverse TIRs of each transposon are very similar; this is also true for the two clustered transposons. From the dendrogram, groups are visible that weren’t found by the VSEARCH clustering; this indicates that, while the TIR sequences are related, these groups likely have different genetic material between their TIR sequences.

## 4.4 Alignment of Transposable Elements

After clusters of transposable elements have been identified, it may be useful to produce an alignment. Since we know that the transposable elements identified by VSEARCH have a minimum 60% sequence similarity, it will be possible to produce good quality sequence alignments. This can be useful for downstream analysis, such as BLAST searching. In this instance, an alignment was done for cluster 4; so an alignment of only two sequences was carried out.

packMatches <- packAlign(
packMatches,
arabidopsisThalianaRefseq,
saveFolder = "tempOutput"
)
## Rognes T, Flouri T, Nichols B, Quince C, Mahe F (2016)
## VSEARCH: a versatile open source tool for metagenomics
## PeerJ 4:e2584 doi: 10.7717/peerj.2584 https://doi.org/10.7717/peerj.2584
##
## vsearch v2.14.1_win_x86_64, 7.9GB RAM, 4 cores
## https://github.com/torognes/vsearch
##
## vsearch v2.14.1_win_x86_64, 7.9GB RAM, 4 cores
## https://github.com/torognes/vsearch
##
## 9396 nt in 6 seqs, min 713, max 2463, avg 1566
## Aligning 100%
## Matching query sequences: 5 of 6 (83.33%)

As with VSEARCH clustering, a .uc and .blast file are created by the alignment. The alignment function returns a summary of the .uc file, however the full details of each file can be recovered as before.

## 4.5 BLAST Analysis

BLAST search has not been implemented in this package, so it is expected that the user will define clusters of transposable elements above before BLAST investigation. This stage will identify the tranposable elements that are autonomous and confirm which matches are likely pack-TYPE elements.

Since this example included only a small subset of the Arabidopsis thaliana genome, and there were very few elements identified, a BLASTN search was carried out for cluster 4. In a larger study, it would have been wise to instead select more interesting clusters and BLAST a subset of these.

# the packMatches dataframe is exported as a FASTA file for NCBI blast search
packsToFasta(
packMatches,
"tempOutput/packMatches.fasta",
arabidopsisThalianaRefseq
)
## [1] "FASTA written to tempOutput/packMatches.fasta"

After running the cluster 4 sequence using NCBI’s BLASTX service, two major hits were found (GENBANK: AAD28056.1 and AAF06087.1) for the sequence. One match appeared to be similar to a putative En/Spm-like transposon protein, suggesting that this transposon may not be a Pack-TYPE element as it may not have captured genetic material from the host genome.

# 5 Data Formats and Conversion

## 5.1 packSearch Dataframe

Dataframes are used by packFinder to contain the locations of predicted elements and additional metadata, such as TSD sequence and cluster designation. The dataframe is in a similar format to that stored by GenomicRanges. The dataframe can be saved and restored, as below, for longer term storage and export of packFinder results. packMatches refers to the object created in previous steps, by applying packSearch followed by packClust.

packsToCsv(packMatches, "tempOutput/packMatches.csv")
## [1] "File successfully written to tempOutput/packMatches.csv"
print(getPacksFromCsv("tempOutput/packMatches.csv"))
##   seqnames   start     end width strand TSD cluster
## 1     Chr3  100830  102347  1518      + TGT       3
## 2     Chr3 1068802 1069514   713      + ATA       4
## 3     Chr3 2807747 2809454  1708      + GGT       2
## 4     Chr3 3487540 3488267   728      + ATA       4
## 5     Chr3 3582297 3584562  2266      + ATA       1
## 6     Chr3 3738747 3741209  2463      + TTG       0

## 5.2 GRanges

As mentioned previously, the packFinder dataframe is in a similar format to that stored by GenomicRanges. The data produced by packFinder can therefore be quickly converted to and from a GRanges object, as below.

packsGRanges <- packsToGRanges(packMatches)
print(packsGRanges)
## GRanges object with 6 ranges and 2 metadata columns:
##       seqnames          ranges strand |         TSD   cluster
##          <Rle>       <IRanges>  <Rle> | <character> <integer>
##   [1]     Chr3   100830-102347      + |         TGT         3
##   [2]     Chr3 1068802-1069514      + |         ATA         4
##   [3]     Chr3 2807747-2809454      + |         GGT         2
##   [4]     Chr3 3487540-3488267      + |         ATA         4
##   [5]     Chr3 3582297-3584562      + |         ATA         1
##   [6]     Chr3 3738747-3741209      + |         TTG         0
##   -------
##   seqinfo: 1 sequence from an unspecified genome; no seqlengths
print(getPacksFromGRanges(packsGRanges))
##   seqnames   start     end width strand TSD cluster
## 1     Chr3  100830  102347  1518      + TGT       3
## 2     Chr3 1068802 1069514   713      + ATA       4
## 3     Chr3 2807747 2809454  1708      + GGT       2
## 4     Chr3 3487540 3488267   728      + ATA       4
## 5     Chr3 3582297 3584562  2266      + ATA       1
## 6     Chr3 3738747 3741209  2463      + TTG       0

If many elements are discovered by packFinder it may be necessary to identify overlapping elements, that could not be produced by a transposase, and remove them; this can be done using GenomicRanges.

## 5.3 FASTA

Additionally, the dataframe produced by packFinder can be exported to and restored from FASTA formats. While this will not save additional metadata columns, such as cluster designations, the core information will be preserved in the FASTA title field.

packsToFasta(
packMatches,
"tempOutput/packMatches.fasta",
arabidopsisThalianaRefseq
)
## [1] "FASTA written to tempOutput/packMatches.fasta"
print(getPacksFromFasta("tempOutput/packMatches.fasta"))
##   seqnames   start     end width strand TSD
## 1     Chr3  100830  102347  1518      + TGT
## 2     Chr3 1068802 1069514   713      + ATA
## 3     Chr3 2807747 2809454  1708      + GGT
## 4     Chr3 3487540 3488267   728      + ATA
## 5     Chr3 3582297 3584562  2266      + ATA
## 6     Chr3 3738747 3741209  2463      + TTG

# 6 Step-wise packSearch Functions

While it is recommended to use packSearch to identify potential Pack-TYPE elements, it is possible to run the individual functions that make up the packSearch pipeline. Below are the steps used by packSearch to predict transposons, using the same parameters as the previous example.

## 6.1 Identifying Potential TIRs

Potential TIR sequences are first identified by pattern matching.

forwardMatches <- identifyTirMatches(
Biostrings::DNAString("CACTACAA"),
arabidopsisThalianaRefseq,
strand = "+",
tsdLength = 3
)
nrow(forwardMatches)
## [1] 90
reverseMatches <- identifyTirMatches(
Biostrings::reverseComplement(Biostrings::DNAString("CACTACAA")),
arabidopsisThalianaRefseq,
strand = "-",
tsdLength = 3
)
nrow(reverseMatches)
## [1] 97

## 6.2 Obtaining TSD Sequences

The function getTsds may be used to quickly obtain the TSD sequences for TIRs. This function can also be used for obtaining TSDs from the ranges of known full transposons, given a dataframe in the format produced by packSearch.

forwardMatches$TSD <- getTsds( forwardMatches, arabidopsisThalianaRefseq, 3, strand = "+" ) head(forwardMatches) ## seqnames start end width strand TSD ## 1 Chr3 100830 100837 8 + TGT ## 2 Chr3 135422 135429 8 + TTA ## 3 Chr3 139547 139554 8 + AAC ## 4 Chr3 146399 146406 8 + AAC ## 5 Chr3 179436 179443 8 + TTT ## 6 Chr3 216837 216844 8 + ATT reverseMatches$TSD <- getTsds(
reverseMatches,
arabidopsisThalianaRefseq,
3,
strand = "-"
)

head(reverseMatches)
##   seqnames  start    end width strand TSD
## 1     Chr3    447    454     8      - AAT
## 2     Chr3  42505  42512     8      - CGT
## 3     Chr3  84518  84525     8      - ATT
## 4     Chr3  95521  95528     8      - TTG
## 5     Chr3 102340 102347     8      - TGT
## 6     Chr3 186733 186740     8      - GTG

## 6.3 Filtering TIR Matches

The main step of the algorithm matches TIR sequences together based on their proximity and TSD sequence.

identifyPotentialPackElements(
forwardMatches,
reverseMatches,
arabidopsisThalianaRefseq,
c(300, 3500)
)
##   seqnames   start     end width strand
## 1     Chr3  100830  102347  1518      *
## 2     Chr3 1068802 1069514   713      *
## 3     Chr3 2807747 2809454  1708      *
## 4     Chr3 3487540 3488267   728      *
## 5     Chr3 3582297 3584562  2266      *
## 6     Chr3 3738747 3741209  2463      *

After attaching the TSD sequences to the dataframe of matches, the output matches that of the previous example run using the packSearch pipeline.

## 6.4 Get Transposon Sequences

The function getSeqs may be used to obtain transposon sequences from a dataframe in the format produced by packSearch. This could additionally be used on other dataframes of ranges, such as that produced by identifyTirMatches above.

getPackSeqs(packMatches, arabidopsisThalianaRefseq)
## DNAStringSet object of length 6:
##     width seq                                               names
## [1]  1518 CACTACAAAAAAAAAGTCGAATT...AAAATGATCACTTTTTTGTAGTG Chr3
## [2]   713 CACTACAAGGACAATGAAATTTA...CGTTTACCTGTTTTCTTGTAGTG Chr3
## [3]  1708 CACTACAAAACAAAGTGGATTAC...TTAAAATGATCTTTTTTGTAGTG Chr3
## [4]   728 CACTACAAGGACAATGAAATTTT...CTTTTACCTGTTTTCTTGTAGTG Chr3
## [5]  2266 CACTACAACAAATATCCACATTC...TAATGTGGATATTTGTTGTAGTG Chr3
## [6]  2463 CACTACAAGAATATTAAACTTTA...GAAAAGTCATTTTTCTTGTAGTG Chr3

# 7 References

Information on the properties of Pack-TYPE transposable elements was obtained from the following papers during the development of packFinder.

• Catoni, M. et al. (2019) ‘Mobilization of Pack-CACTA transposons in Arabidopsis suggests the mechanism of gene shuffling’, Nucleic Acids Research. Oxford University Press, 47(3), pp. 1311–1320. doi: 10.1093/nar/gky1196.
• Jiang, N. et al. (2004) ‘Pack-MULE transposable elements mediate gene evolution in plants’, Nature, 431(7008), pp. 569–573. doi: 10.1038/nature02953.

The Bioconductor packages GenomicRanges and Biostrings are used for detection of tranposable elements, and manipulation of DNA sequences.

# 8 Session Information

This vignette was compiled during the following session:

sessionInfo()
## R version 4.2.0 RC (2022-04-19 r82224)
## Platform: x86_64-pc-linux-gnu (64-bit)
## Running under: Ubuntu 20.04.4 LTS
##
## Matrix products: default
## BLAS:   /home/biocbuild/bbs-3.15-bioc/R/lib/libRblas.so
## LAPACK: /home/biocbuild/bbs-3.15-bioc/R/lib/libRlapack.so
##
## locale:
##  [1] LC_CTYPE=en_US.UTF-8       LC_NUMERIC=C
##  [3] LC_TIME=en_GB              LC_COLLATE=C
##  [5] LC_MONETARY=en_US.UTF-8    LC_MESSAGES=en_US.UTF-8
##  [7] LC_PAPER=en_US.UTF-8       LC_NAME=C
## [11] LC_MEASUREMENT=en_US.UTF-8 LC_IDENTIFICATION=C
##
## attached base packages:
## [1] stats     graphics  grDevices utils     datasets  methods   base
##
## other attached packages:
## [1] packFinder_1.8.0 BiocStyle_2.24.0
##
## loaded via a namespace (and not attached):
##  [1] Rcpp_1.0.8.3           highr_0.9              bslib_0.3.1
##  [4] compiler_4.2.0         BiocManager_1.30.17    kmer_1.1.2
##  [7] jquerylib_0.1.4        GenomeInfoDb_1.32.0    XVector_0.36.0
## [10] bitops_1.0-7           tools_4.2.0            zlibbioc_1.42.0
## [13] digest_0.6.29          nlme_3.1-157           jsonlite_1.8.0
## [16] evaluate_0.15          lattice_0.20-45        rlang_1.0.2
## [19] cli_3.3.0              magick_2.7.3           parallel_4.2.0
## [22] yaml_2.3.5             phylogram_2.1.0        xfun_0.30
## [25] fastmap_1.1.0          GenomeInfoDbData_1.2.8 stringr_1.4.0
## [28] knitr_1.38             Biostrings_2.64.0      sass_0.4.1
## [31] S4Vectors_0.34.0       IRanges_2.30.0         grid_4.2.0
## [34] stats4_4.2.0           R6_2.5.1               rmarkdown_2.14
## [37] bookdown_0.26          magrittr_2.0.3         htmltools_0.5.2
## [40] BiocGenerics_0.42.0    GenomicRanges_1.48.0   ape_5.6-2
## [43] stringi_1.7.6          RCurl_1.98-1.6         crayon_1.5.1