VGEA: an RNA viral assembly toolkit

Introduction

The most abundant biological entities on Earth are viruses as they can be found among all cellular forms of life. So far, over four thousand five hundred viral species have been discovered, from which a huge amount of sequence information has been collected by researchers and scientists all over the world (Pickett et al., 2012; Sharma, Priyadarshini & Vrati, 2015; Brister et al., 2015). In recent times (past two decades), a number of these viruses have emerged in the human population causing disease outbreaks and sometimes pandemics. These viruses include mainly: Influenza virus, Severe Acute Respiratory Syndrome (SARS) coronavirus, Middle East Respiratory Syndrome (MERS) coronavirus, Ebola virus, Yellow fever virus, Lassa virus (LASV), Zika virus (Chan, 2002; Bean et al., 2013; Folarin et al., 2016; Grubaugh et al., 2017; Metsky et al., 2017; Siddle et al., 2018; Ajogbasile et al., 2020) and SARS-CoV-2 (Chen et al., 2020; Holshue et al., 2020; Sohrabi et al., 2020). During these outbreaks and pandemics, genomic sequencing for identification and characterization of the transmission and evolution of the causative agents have proved to be critical in helping inform disease surveillance and epidemiology.

Next Generation Sequencing (NGS) platforms have been widely accepted as high-throughput, open view technologies that have many attractive features for virus detection and assembly (Tang & Chiu, 2010; Mokili, Rohwer & Dutilh, 2012). NGS-based studies have vastly increased our understanding of viral diversity (Reyes et al., 2010; Cantalupo et al., 2011). Pathogen sequence data obtained from NGS experiments are a rich source of information, these data can be used to study their epidemiology, evolution, transmission patterns, and can also inform drug and vaccine design. The field of genomics, especially pathogen genomics has been transformed by NGS, with costs constantly decreasing, equipment becoming more portable/field deployable during outbreaks and remarkable increase in data availability.

The huge amount of data being generated requires various processing steps such as removal of primers and adapters, quality filtering and control which is usually crucial for various downstream analysis. Several tools have been developed for these purposes, such as fastp (Chen et al., 2018) and Trimmomatic (Bolger, Lohse & Usadel, 2014).

Reconstructing viral genomes from NGS data is usually achieved through de novo assembly (which is the process of assembling genomes using overlapping sequencing reads), or through a reference-guided approach (which involves mapping sequence reads to a reference genome). Numerous tools have been developed for these purposes; SPAdes (Bankevich et al., 2012), Burrows-Wheeler Alignment tool (BWA), V-GAP (Nakamura et al., 2016), VirusTAP (Yamashita, Sekizuka & Kuroda, 2016), V-Pipe (Posada-Céspedes et al., 2021) and viral-ngs (https://github.com/broadinstitute/viral-ngs), amongst others. Contigs generated by de novo assembly however do not provide a complete summary of reads, misassembly can result in the contigs having an incorrect structure, and for parts of the genome where contigs could not be assembled, no information is available. In addition, reference-guided assembly of viral genomes can lead to biased loss of information which can then skew epidemiological and evolutionary conclusions (Wymant et al., 2018).

Variant analysis and genome quality assessment to detect variants and changes occurring across the genome of a virus is also a key step in viral genome analysis as viruses (especially RNA viruses) are known to have high mutation rates (Duffy, 2018). Variant analysis is important for detecting outbreak origins and for phylogenetic/phylogeographic studies and best practices for variant identification in microbial genomes have been proposed in literature and adopted to a large extent (Van der Auwera et al., 2013).

A number of pipelines that have been developed for downstream analysis of viral genomes require high performance computing (HPC) clusters and/or cloud-based systems e.g., the V-pipe authors recommend running V-pipe on clusters because for most applications, running V-pipe on a local machine may not be efficient (https://github.com/cbg-ethz/V-pipe/wiki/advanced) and some of these pipelines are only web-based such as VirAmp (Wan et al., 2015) and VirusTAP (Yamashita, Sekizuka & Kuroda, 2016. Also, some pipelines have many dependencies to be installed especially if the analysis requires multiple tasks to be performed. In low-and-middle income countries (LMICs) where most scientists do not have access to HPC clusters or cloud-based systems and where internet connection is too unstable to regularly make use of web-based platforms for analysis, this can be a daunting task.

The challenges listed above motivated the development of VGEA (Viral Genomes Easily Analyzed, available online at https://github.com/pauloluniyi/VGEA). VGEA makes use of existing bioinformatics pipeline/tools to carry out various viral genome analysis tasks and is built on an advanced workflow management system, Snakemake (Köster & Rahmann, 2012).

Materials and Methods

Implementation

The installation of VGEA requires the pipeline to be downloaded onto a personal computer and creation of a conda environment to set up all dependencies. Complete installation steps are in the github README file: https://github.com/pauloluniyi/VGEA/blob/master/README.md

The analysis of VGEA is broken down into a set of ‘rules’ that links the output file of an analysis into the input of the next task in the general workflow (Fig. 1). The dependencies are fastp for read trimming and read-level quality control, BWA for mapping sequencing reads to the human reference genome, SAMtools for extracting unmapped reads and also for splitting bam files into fastq files, IVA for de novo assembly to generate contigs, shiver to pre-process reads for quality and contamination, then map to a reference tailored to the sample using corrected contigs supplemented with the user’s choice of existing reference sequences, SeqKit for cleaning shiver assembly for QUAST, QUAST to evaluate/assess the quality of genome assemblies and MultiQC for aggregation of the results from fastp, BWA and QUAST

All of these tools can be installed using a bioconda channel (Grüning et al., 2018). The input files for VGEA are paired-end fastq files. VGEA allows full customization of the pipeline, so users can modify the parameters used in running their samples. It is possible to modify every step of the workflow to suit the samples being processed. Users can also add more steps to the pipeline as they see fit. The pipeline runs on Linux/Unix and Mac. However, no prior programming is required to run the pipeline and, once the user supplies the input, the whole workflow can run automatically from beginning to end.

Results

VGEA carries out read trimming and quality control tasks on input FASTQ data using fastp (Fig. 2). This increases the quality of data used for subsequent steps of the pipeline.

VGEA then maps reads to the human reference genome in order to remove human contaminants, the pipeline carries out this step using BWA. Genome assembly and consensus sequence generation is carried out, together with the generation of summary minority-variant information (base frequencies at each position) and detailed minority-variant information (all reads aligned to their correct position in the genome). VGEA carries out assembly using IVA and generates consensus sequences using shiver. Previous study by the shiver developers has shown the systematic superiority of mapping to shiver’s constructed reference compared with mapping the same reads to the closest of 3,249 references: median values of 13 bases called differently and more accurately, zero bases called differently and less accurately, and 205 bases of missing sequence recovered (Wymant et al., 2018).

VGEA also assesses the quality of genome assemblies using QUAST. QUAST evaluates metrics such as contig sizes, misassemblies and structural variations, genome representation and its functional elements, variations of N50 based on aligned blocks and then presents these statistics in graphical form. QUAST also makes a histogram of several metrics including the number of complete genes, operons and the genome fraction (%). Finally, VGEA compiles the results of BWA, fastp and QUAST into a single MultiQC report (Fig. 3).

Discussion

VGEA is built on the snakemake workflow management system (BKöster & Rahmann, 2012), a workflow management system that allows the effortless deployment and execution of complex distributed computational workflows in any UNIX-based system, from local machines to high-performance computing clusters. It is a user-friendly, customizable and reproducible pipeline which can be deployed on a personal computer and which can run from start to finish with a single command.

VGEA was designed with ease-of-use in mind and so all its dependencies can be installed in a conda environment under the bioconda channel (Grüning et al., 2018 making it particularly useful for scientists with little or no computational background and for scientists in LMICs who don’t have much access to high-performance computing clusters or cloud-computing resources. VGEA capitalizes on Snakemake’s multi-threading feature so that makes it possible for it to be deployed on laptops with greater computing performance or a computing server to improve its speed. The pipeline was tested with paired-end short-read sequencing data produced by the illumina platform (MiSeq, MiSeq FGx and HiSeq 2500).

The results generated by the major steps of the VGEA pipeline are summed up together into a MultiQC report which can be easily interpreted and understood by anyone with little or no knowledge of bioinformatics.

Conclusion

VGEA was built primarily by biologists and in a manner that is easy to be employed by users without significant computational background. As new and innovative tools for viral genome analysis and assembly are increasingly being developed, these can easily be incorporated into the VGEA pipeline. We hope that other scientists can build upon and improve VGEA as a tool to extract more qualitative and quantitative information from viral genomes.

Supplemental Information