CanVAS: A Harmonized and Imputed Canine Variant Atlas1

The domestic dog (Canis lupus familiaris) is a powerful model for genetic studies of complex disease, but canine genotype data are distributed across independent studies using incompatible genotyping platforms, genome builds, strand conventions, and allele coding schemes. Here we present CanVAS, a quality-controlled, harmonized, and imputed canine genotype resource integrating 15 publicly available datasets into a single analysis-ready PLINK file set on the CanFam4 (UU_Cfam_GSD_1.0) reference assembly. The typed backbone contains 15,451 dogs from over 375 breeds, village dog populations, dingoes, wolves, and coyotes, genotyped across 77,215 shared SNPs. Imputation against the Dog10K whole-genome sequencing reference panel (1,929 dogs) using Beagle 5.4 expanded the resource to 9.7 million variants (DR2 [≥] 0.3, MAF [≥] 0.01), including approximately 3 million rare variants (MAF < 0.05). We describe the complete harmonization pipeline and validate the resource through population structure analysis and genome-wide runs-of-homozygosity analysis, recovering known breed-level differences in genomic inbreeding.