Skiver: Reference-free quality control of metagenomic sequencing datasets using (k,v)-mer sketches

Quality control is essential in numerous Bioinformatics pipelines, yet existing approaches often depend on inefficient mapping to reference genomes or potentially uncalibrated quality scores. We present skiver, a reference-free, alignment-free quality control tool that uses (k, v)-mer sketches. By deriving consensus from sketched (k, v)-mers, skiver estimates survival and hazard rates that capture the positional distribution of errors. Across simulated and real datasets from multiple sequencing platforms, skiver accurately estimates overall error rates, error-type proportions, and quality-score calibration. An iterative outlier-filtering strategy enables robust performance on complex datasets with multiple strains, alleles, and repetitive regions.