UniqSketch: sensitive and resource-efficient strain-level detection in metagenomes

Microbial communities contain diverse organisms, including multiple coexisting strains of the same species that display phenotypic differences influencing pathogenicity and host interactions. Detecting strains is challenging because of computational demands and their low abundance in metagenomes. We present UniqSketch, an open-source toolkit for identifying target strains across species within a single metagenome. It tracks specific microbial signatures or contaminants even in complex, multi-species samples. UniqSketch uses unique genomic fingerprints and a compact Bloom filter for efficient computation and accurate abundance estimation. Results show improved precision and accuracy over existing tools, with reliable performance at sequencing coverage as low as 0.1×.