Chromosome-level haplotype-resolved assembly of highly heterozygous grass genomes with PhaseGrass

Abstract Current methods to generate haplotype-resolved assemblies for highly heterozygous genomes suffer from reference bias when relying on reference-based phasing or may generate imbalanced haplomes when relying on graph-based phasing. Here, we report PhaseGrass, an assembly, phasing and scaffolding workflow that is compatible with accurate or error-prone long reads, generating haplotype-resolved assemblies for highly heterozygous genomes. Requiring no parental data, PhaseGrass combines reference-based phasing with haplotype-specific k-mers to partition long reads or unitigs to corresponding haplotypes, thereby avoiding reference bias and uneven haplotype partitioning. Using PhaseGrass, we generated the first chromosome-level, haplotype-resolved assemblies for the highly heterozygous, allogamous grass species Lolium perenne and L. multiflorum. We found that PhaseGrass binned 20% more reads to haplotypes than WhatsHap and generated balanced haplomes compared to Hifiasm, which produced largely imbalanced haplomes for L. multiflorum due to abundant structural variations between haplotypes. PhaseGrass will facilitate routine generation of haplotype-resolved pangenomes for species with highly heterozygous genomes.