Answers are grounded in this paper and cite their sources.
Existing read phasing tools operate on reads aligned to a linear reference, where reference bias reduces the information available for haplotype inference and read assignment. Pangenome graphs reduce reference bias, but reads falling within non-reference sequence cannot be surjected to a linear reference, so phasing after surjection reintroduces that bias. Here we present GRASER, which phases graph-aligned short reads against haplotype paths inferred for each individual, using established graph genotyping and statistical phasing to build those paths. GRASER phase uses edges unique to one haplotype path as evidence, so reads are phased directly from graph alignments without conversion to a linear reference. Against true parental haplotype paths, GRASER assigned 22.6 to 24.5% of simulated fragments to a haplotype, and more than 96% of those were placed on the correct haplotype. Inferred haplotype paths reduced the assignment rate by about 4 percentage points. GRASER assigned more fragments than a WhatsHap-based linear-reference workflow in every configuration tested. Applied to whole-genome bisulfite sequencing, GRASER recovered allele-specific methylation at imprinted loci and measured haplotype-resolved methylation within a structural variant.
No takes yet. Share an insight, caveat, or question.
Garza et al. (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: