TranspoScope: interactive visualization of retrotransposon insertions

Abstract

Motivation

Retrotransposition is an important force in shaping the human genome and is involved in prenatal development, disease and aging. Current genome browsers are not optimized for visualizing the experimental evidence for retrotransposon insertions.

Results

We have developed a specialized browser to visualize the evidence for retrotransposon insertions for both targeted and whole-genome sequencing data.

Availability and implementation

TranspoScope’s source code, as well as installation instructions, are available at https://github.com/FenyoLab/transposcope.

1 Introduction

Long interspersed elements (LINEs) are a group of retrotransposons (sequences that copy themselves to new genomic loci via an RNA intermediate), which make up ∼17% of the human genome (Ostertag and Kazazian, 2001). Although LINE-1 is the only family of retrotransposons that is still active in humans, most of the roughly 500 000 copies (Lander et al., 2001) are no longer capable of retrotransposition due to 5′ truncation during integration or to acquired mutations (Brouha et al., 2003). However, a small number of ‘hot LINE-1’ elements from the L1Hs subfamily are still capable of retrotransposition and are responsible for most L1 retrotransposition in humans (Brouha et al., 2003).

Various computational tools targeting different sequencing strategies exist for detecting LINE-1, such as the Mobile Element Locator Tool (MELT) (Gardner et al., 2017) for whole-genome sequencing data and TIPseqHunter (Tang et al., 2017) for Transposon Insertion Profiling by Sequencing (TIPseq) data. TIPseq utilizes vectorette PCR to selectively amplify DNA flanking the 3′ end of LINE-1 elements (Steranka et al., 2019).

As de novo and polymorphic LINE-1 insertions within a given individual will not be reflected within the reference genome, reads which overlap with these specific insertions will not be aligned correctly. As numerous copies of pre-existing LINE-1 are present in the reference genome, reads that should align to the de novo LINE-1 insertion are likely to be misaligned to one of the existing copies. This makes visualizing read pairs impractical as they could potentially be mapped to incorrect loci within the genome.

To solve this problem, we developed a novel tool that creates a localized reference sequence incorporating the relevant sequence from LINE-1 as well as the sequence flanking the LINE-1 insertion from the reference genome. Reads flanking the LINE-1 insertion are realigned to the localized reference to determine whether the predicted insertion truly exists in this location. The read coverage from the local realignment can then be viewed within a web-based visualization tool described below.

2 Implementation

The TranspoScope Package comprises two steps. First, local realignment for each potential insertion is performed. Second, an interactive web-based visualization is used to view the configuration of reads relative to each insertion site. The TranspoScope package enabling these processes has been made available on Bioconda, a software repository designed to normalize the software installation process regardless of the operating system (Grüning et al., 2018).

2.1 Local realignment

The code used for performing the local realignment is written in Python. For each potential insertion site, reads which fall within a given flanking region of the site are extracted. These reads are then used to gather their paired ends which may map to different locations within the genome. Read pairs are extracted even if they are aligned discordantly. The gathered read pairs are then written to FASTQ format files for each insertion site. Second, a local realignment is then performed using Bowtie2 (Langmead and Salzberg, 2012) for each insertion. The combined reference sequence includes two different DNA fragments. One is the region of L1Hs which has been identified using an appropriate tool (MELT, TIPseqHunter). The other is extracted from the human reference genome which is provided as an argument when running TranspoScope using either a left, right or both flanking regions based on the estimated strand information and sequencing technology. The new aligned CRAM file is then used to estimate coverage profiling in the coverage viewer.

2.2 Interactive coverage viewer

The TranspoScope viewer is served as a static webpage to simplify the process of publishing data online. This is possible as the CRAM files (described in Section 2.1) offer a compact representation of the read data that can be parsed in the browser using cram-js (Buels et al., 2016). A Python-based web server is provided for viewing the output locally and can be started using the command line. As the output from the realignment step is a folder containing the files required to serve the static web-page, this folder can be hosted online using any web server.

Figure 1 depicts an example from the visualization tool we developed which displays several stacked histograms that outline different read pair configurations relative to a LINE-1 insertion which was identified using MELT on WGS data.

Fig. 1.

TranspoScope web-based visualization. The visualization highlights different read pair configurations through the use of colored histograms. The light blue vertical rectangles indicate the regions which contain target site duplications at the 5′ and 3′ end of the LINE-1 insertion. Gray histograms represent read pairs in which both reads align to the reference genome. Orange histograms represent read pairs where one or both of the reads overlap the junction region. Green histograms indicate read pairs where one read is unmapped. Red histograms indicate read pairs where one read is located on each side of the LINE-1 sequence. Finally, purple and blue histograms indicate reads where one read aligns to the reference genome (purple) and the other aligns with the LINE-1 sequence (blue)

Additional features of TranspoScope include the ability select specific histograms, an option to display the specific sequences of reads which cross the junction region(s), the ability to zoom into the region surrounding the mouse cursor down to a single base pair resolution and the ability to view base pair statistics within a pop-up window by hovering over a given histogram.

3 Conclusions

TranspoScope provides an effective means for visualizing the manner in which read pairs support potential LINE-1 insertions. In addition, TranspoScope allows users to visually confirm the sequences of individual reads that overlap the junction region(s). TranspoScope has been used to visualize potential insertions identified by TIPseqHunter within TIPseq experiments as well as those identified by MELT found within WGS data. When utilizing WGS data, it is possible to visualize both the 5′ as well as the 3′ end of an insertion using TranspoScope. TranspoScope has proven effective in visualizing the LINE-1 retrotransposon but with minor adjustments, the software will be capable of visualizing any retrotransposon as well as genetic rearrangements.