Gfa2bin enables graph-based GWAS by converting genome graphs to pan-genomic genotypes

Variation graphs offer superior representation of genomic diversity compared to traditional linear reference genomes, capturing complex features, especially structural variants, that are otherwise inaccessible to analysis. Given the known role of structural variants in functional allelic variation, it seems self-evident that exploiting variation graphs for genome-wide association studies (GWAS) should enable more comprehensive knowledge of associations between genetic variants and phenotypic traits. We present a tool, gfa2bin, that offers multiple methods to (i) use variation graphs for genotyping, and (ii) convert the genotypes into well-established data formats for use in GWAS. In a case study where we apply gfa2bin to Arabidopsis thaliana, we demonstrate with 1,695 traits that variation graphs can detect genotype-phenotype associations not found with linear reference genomes. Our approach complements SNP-based methods; not only does graph-based GWAS produce additional associations, but associations also found with SNPs have on average higher significance in graph-based GWAS. gfa2bin is implemented in Rust, and commented source code is available under an MIT license at https://github.com/MoinSebi/gfa2bin. Examples of how to run gfa2bin are provided in the documentation. We provide several Python scripts and a Snakemake pipeline for easy use of the tool with larger data sets. The packing tool (https://github.com/MoinSebi/packing) is recommended for reduced storage and preprocessing (normalization) of sequence-to-graph alignments.