BiTUGA: scalable prevalence-based unitig association testing for binary traits

Sequence-based association studies can be challenging in large and highly repetitive genomes. While reference-free k-mer approaches enable direct analysis of sequence variation from sequencing reads, current tools often rely on constructing full k-mer-by-sample matrices, creating computational bottlenecks. Here, we present BiTUGA, a pipeline to test associations for discrete binary traits designed to overcome these limitations and optimized for large genomes. Shifting the statistical focus from raw abundance to group-level prevalence, BiTUGA assembles filtered k-mers into unitigs and tests unitig presence for association across groups. We validated BiTUGA on the binary trait sex, targeting structurally complex Sex-Determining Regions (SDRs) that remain largely unresolved in many plant species. BiTUGA identified sex-associated unitigs, detecting the known SDRs in Populus tremula and Ginkgo biloba, processing up to 750 Gbp in 14-25 h within 70 GB RAM. BiTUGA is available at https://github.com/JMittelbach/BiTUGA.git.