ViR: a tool to solve intrasample variability in the prediction of viral integration sites using whole genome sequencing data

Abstract

Background Several bioinformatics pipelines have been developed to detect sequences from viruses that integrate into the human genome because of the health relevance of these integrations, such as in the persistence of viral infection and/or in generating genotoxic effects, often progressing into cancer. Recent genomics and metagenomics analyses have shown that viruses also integrate into the genome of non-model organisms (i.e., arthropods, fish, plants, vertebrates). However, rarely studies of endogenous viral elements (EVEs) in non-model organisms have gone beyond their characterization from reference genome assemblies. In non-model organisms, we lack a thorough understanding of the widespread occurrence of EVEs and their biological relevance, apart from sporadic cases which nevertheless point to significant roles of EVEs in immunity and regulation of expression. The concomitance of repetitive DNA, duplications and/or assembly fragmentations in a genome sequence and intrasample variability in whole-genome sequencing (WGS) data could determine misalignments when mapping data to a genome assembly. This phenomenon hinders our ability to properly identify integration sites. Results To fill this gap, we developed ViR, a pipeline which solves the dispersion of reads due to intrasample variability in sequencing data from both single and pooled DNA samples thus ameliorating the detection of integration sites. We tested ViR to work with both in silico and real sequencing data from a non-model organism, the arboviral vector Aedes albopictus. Potential viral integrations predicted by ViR were molecularly validated supporting the accuracy of ViR results. Conclusion ViR will open new venues to explore the biology of EVEs, especially in non-model organisms. Importantly, while we generated ViR with the identification of EVEs in mind, its application can be extended to detect any lateral transfer event providing an ad-hoc sequence to interrogate.

High-throughput detection of eukaryotic parasites and arboviruses in mosquitoes

“…2 ) and, based on their phylogenic position, likely to infect mosquitoes rather than humans. Alternatively, these sequences could represent polymorphic viral integrations into the mosquito genome ( Palatini et al, 2017 ; Whitfield et al, 2017 ; Russo et al, 2019 ; Pischedda et al, 2021 ).…”

“…We identified several viruses, distinct from known viruses (Figure 2) and, based on their phylogenic position, likely to infect mosquitoes rather than humans. Alternatively, these sequences could represent polymorphic viral integrations into the mosquito genome (48)(49)(50)(51).…”

Computational methods for the discovery and annotation of viral integrations

“…Finally, a script refines the output including the final virus integration candidates. ViR works downstream of Vy-PER, or any other paired end reads-based EVE prediction algorithm, to improve predictions of viral integration sites by addressing the dispersion of reads due to intrasample variability (Figure 2) [13]. Intrasample variability can be due to repetitive DNA and/or fragmentation in the assembly.…”

“…The most important factors influencing the computational power, running time and disk occupancy required by the procedures described here are 1) the size of the reference genome assembly, 2) the size and number of the raw sequencing reads (fastq) files and the consequent aligned reads (sam/bam) files, and 3) the size of the databases (of viral proteins and/or genomes) used to search for EVEs. A high-performance computing (HPC) cluster or a dedicated server with at least 32 computing threads and 64GB of RAM are recommended, especially to run Vy-PER as this pipeline can run multiple alignment processes in parallel via a scheduling manager [12,13]. The advantage of running multiple alignments in parallel depends on the behavior of BWA, the most computationally-intensive step in the protocol.…”

“…Tasks 2 and 3 are implemented with WGS short reads that are mapped to a genome assembly. Task 3 is executed by running Vy-PER [12], a pipeline originally designed to identify viral integrations into the human genome, followed by ViR [13], a pipeline designed to improve predictions of new integration sites by accounting for dispersion of reads in repetitive DNA sequences, such as piRNA clusters. The identification of viral integrations using alignment of WGS data to a genome assembly is based on the identification of chimeric reads (a read pair in which one read maps to the host genome, hereafter referred to as host read, and the other read to a virus, hereafter referred to as viral read) and/or unmapped reads, which are extracted from the WGS dataset.…”