Impact of GenomeScope assumptions on estimate accuracy

Determine the extent to which deviations from the assumptions of the GenomeScope and GenomeScope 2.0 k-mer spectrum models—including non-uniform distribution of heterozygosity across the genome, the presence of variant types beyond single-nucleotide polymorphisms, and genomic regions with more than two copies—impact the accuracy of parameter estimates produced by these models, and ascertain how this impact varies among species.

Background

GenomeScope and GenomeScope 2.0 infer genome properties (notably heterozygosity and repeat content) by fitting probabilistic models to k-mer spectra. These models assume, among other things, uniformly distributed heterozygosity, that variants are primarily SNPs, and that repetitiveness predominantly involves two-copy regions.

The authors note that real genomes can violate these assumptions—for example, heterozygosity may be non-uniform, many variants are not SNPs, and repeats may have more than two copies—potentially biasing estimates. The magnitude of this bias and its dependence on species remains unresolved, motivating a focused investigation into the robustness of GenomeScope-derived estimates under realistic genomic conditions.