A general and extensible algorithmic framework to biological sequence alignment across scales and applications

Abstract Sequence alignment underpins nearly every facet of modern genomics, from genetic testing and cancer profiling to functional genome annotation. Yet, despite decades of algorithmic innovation, most existing aligners remain narrowly optimized for specific tasks, fragmenting analytical workflows and limiting reproducibility. Here we introduce the Versatile Alignment Toolkit (VAT), a unified algorithmic framework that generalizes existing seeding and genome-indexing strategies within a single, transparent architecture. VAT employs a novel multi-view indexing scheme that integrates multiple seeding strategies and supports run-time seed-length parameterization without reindexing. A radix clustering algorithm based on hardware-efficient in-register bitonic sort accelerates multi-view table construction, ensuring scalability across large datasets. VAT delivers consistently high performance across diverse alignment tasks, including short- and long-read mapping, homology search, and whole-genome alignment, while maintaining algorithmic simplicity and adaptability. By bridging previously isolated alignment paradigms, VAT substantially reduces workflow complexity, enhances computational efficiency, and establishes an extensible foundation for future sequencing technologies. We anticipate this unification will set a new standard for flexible and reproducible sequence alignment in biomedical research and clinical genomics.