Using pangenome variation graphs to improve mutation detection in a large DNA virus

Abstract

Accurately quantifying viral genetic diversity is essential for understanding pathogen evolution, transmission, and emergence. However, standard approaches that map sequencing reads to a single linear reference genome introduce substantial reference bias, particularly for samples that are divergent, recombinant, or belong to rare lineages. Pangenome variation graphs (PVGs) mitigate this issue by representing multiple genomes within a unified graph structure, enabling read mapping across all observed and potential haplotypes. Despite this, PVGs have rarely been applied to viruses. Here, we address this gap by constructing and evaluating the first PVG for lumpy skin disease virus (LSDV), an emerging poxvirus of global importance. We generated PVGs of different sizes and mapped Illumina datasets using Giraffe, benchmarking performance against linear reference mapping with Minimap2. A minimal three-sample PVG containing one representative from each major lineage recovered 97% of known LSDV nucleotide diversity while reducing PVG size by >95% relative to a 121-sample PVG. PVG-based mapping detected more SNPs than linear mapping, including variants supported by read evidence that was not detected when reads were mapped to a single reference genome. 27% of SNPs detected using PVG-based mapping could not be projected onto the linear reference coordinate system because they were on alternative paths absent from reference, highlighting the impacts of reference bias. Notably, these new SNPs were at genes involved in host recognition and immune evasion, and identified lineage-specific mutations that improved subclade phylogenetic structure. Our findings demonstrate that PVGs substantially enhance SNP discovery in LSDV, with direct implications for genomic surveillance, outbreak tracing, and the detection of recombinant vaccine-related lineages in LSDV and other large DNA viruses. Impact StatementGenomic surveillance of viral pathogens typically relies on mapping reads to a single reference genome, a practice that systematically misses variation in divergent or recombinant isolates. We show that pangenome variation graphs (PVGs) can reduce this limitation in DNA viruses. Using lumpy skin disease virus (LSDV) as a model system, we demonstrate that a compact PVG constructed from only three representative genomes captures most known genomic diversity and improves the detection of mutations that are difficult to identify using linear reference mapping. PVGs recover biologically meaningful variants at genes involved in immune evasion and host interaction, resolve closely related subclades, and reveal mutations that cannot be identified using a single reference genome. This work uses a population-structure-guided strategy for building reference viral PVGs that balance computational efficiency with sensitivity, offering clear benefits for genomic surveillance, outbreak reconstruction, and the detection of recombinant lineages. Data SummaryAll supporting code and protocols have been provided within the article or through supplementary data files. Supplementary tables are available at Figshare: https://figshare.com/s/7a879df5b5452d04585b with all code at https://github.com/downingtim/LSDV_pangenomics