polars-bio

Genomic interval operations can be performed out-of-core on large BED files by using scan functions for streaming and LazyFrame-first execution to compute overlaps, coverage, and nearest neighbors without loading everything into memory.

VCF and BAM files can be registered and queried using DataFusion SQL, returning results as a LazyFrame to enable further interval operations and seamless bioinformatics file I/O.

Computing nearest intervals, merges, complements, and subtract operations is handled directly on Polars DataFrames and LazyFrames using a DataFrame-centric API, providing fast interval arithmetic for BED-like workflows.

Streaming and scalable I/O are supported for cloud-native genomic reads, utilizing out-of-core processing with predicate and projection pushdown when supported by the underlying bioinformatics file formats.

Coordinate-system handling is explicitly supported for genomic intervals, though format-specific readers and writers for GFF and GTF files may have limitations depending on the streaming predicates and cloud storage paths used.

Polars is required as the core environment because polars-bio provides a Polars DataFrame-centric API, executing interval operations and bioinformatics file I/O directly on DataFrames and LazyFrames.