Format matrix

Oxbow currently supports the following NGS formats, with the encodings and access patterns described in the table below. Oxbow enables random access to most uncompressed or structured, block-compressed data sources and can use conventional format-specific indexes to support efficient genomic range queries.

Format	Modality	Encoding	Multiple resolutions	Compression support[1]	Index support	Authority
FASTQ	Sequence	Text	No	BGZF
FASTA	Sequence	Text	No	BGZF	FAI, FAI + GZI[2]
SAM	Alignment	Text	No	BGZF	TBI, CSI	[3]
BAM	Alignment	Binary	No	BGZF	BAI, CSI	[3]
CRAM	Alignment	Binary	No	many[6]	CRAI	[3]
VCF	Variant call	Text	No	BGZF	TBI, CSI	[3]
BCF	Variant call	Binary	No	BGZF	CSI	[3]
GTF	Feature	Text	No	BGZF	TBI, CSI
GFF3	Feature	Text	No	BGZF	TBI, CSI	[5]
BED	Feature	Text	No	BGZF	TBI, CSI	[3], [4]
BigBed	Feature	Binary	Yes	deflate	Internal index	[4]
BigWig	Feature	Binary	Yes	deflate	Internal index	[4]

---

[1]

BGZF is a backwards-compatible “blocked” variant of GZIP that allows for random access and indexing using one of the companion index files listed in the next column.

[2]

FAI indexes uncompressed FASTA records. GZI is an additional file required to perform look ups in BGZF-compressed FASTA.

[3]

GA4GH and the htslib references and specifications.

[4]

UCSC Genome Browser and the Kent Source.

[5]

sequenceontology.org

[6]

CRAM is a hierarchical format, organized into “containers” and “slices”, with slices being the unit of random access. Each slice consists of data “blocks” that support a number of different encodings, including lossy ones. For example, for CRAM files using reference-based compression, the exact (external) reference sequences are required to decode sequence bases from alignment records.