polars-bio Genomic Intervals

What it does

A reference skill that lets Claude write correct, idiomatic polars-bio code (Polars + Arrow + DataFusion).

• 8 interval operations: overlap, count_overlaps, nearest, merge, cluster, coverage, complement, subtract
• Bioinformatics I/O: BED, VCF (and VCF Zarr), BAM, CRAM, SAM, GFF/GTF, FASTA, FASTQ, Hi-C pairs via read_* / scan_* / write_* / sink_*
• SQL layer: register files as tables and query them with DataFusion SQL
• Pileup/depth: CIGAR-aware per-base read depth from BAM/CRAM
• Streaming & cloud: out-of-core processing plus direct s3://, gs://, az:// reads

It also encodes the tricky parts — 1-based vs 0-based coordinate metadata, .pb only exposing interval ops on LazyFrame, probe-build argument order, INT32 coordinate limits — which sharply reduces hallucinated API calls.

Who it's for

• Genomics analysts hitting performance walls with bioframe or pybedtools
• Researchers processing large VCF/BAM files on memory-limited machines
• Data engineers wiring interval arithmetic into Polars pipelines
• Teams keeping sequencing data in cloud buckets who want query-in-place

Example uses

1. Peak-to-gene annotation: "Annotate my ATAC-seq peaks with the nearest GENCODE gene" → pb.read_bed + pb.read_gff + pb.nearest.
2. Coverage report: "Compute mean depth of aligned.bam across my target panel BED" → pb.depth(min_mapping_quality=20) joined back to intervals.
3. Cloud variant filtering: "Pull QUAL>30 variants from s3://bucket/cohort.vcf.gz and write Parquet" → pb.register_vcf + pb.sql + streaming collect.

Install the skill I found on Claude Skill Mart.
Copy the skills/polars-bio folder from the GitHub repo K-Dense-AI/scientific-agent-skills into my ~/.claude/skills/polars-bio/.
When it's done, tell me in one line what this skill can do.