GTF / GFF to BED

GTF/GFF use 1-based, fully closed coordinates while BED is 0-based with an exclusive end: the start is reduced by one and the end kept, so lengths are unchanged.

“Genes” take gene / pseudogene rows; “transcripts” take transcript rows in GTF and transcript types such as mRNA, tRNA and rRNA in GFF3; other features match the column 3 type exactly.

BED12 gives one row per transcript with exons as blocks and the CDS range in thickStart/thickEnd (both equal the start for non-coding transcripts), ready for display of exon structure in a genome browser.

In GFF3 the gene name comes from the Name (or gene) attribute of the top-level gene; missing names fall back to the gene ID and then the transcript ID. The histogram shows output lengths (log10).

Method

Parses GTF2.2 (Ensembl / Brent lab specification) and GFF3 (Sequence Ontology specification 1.26); BED12 follows the UCSC Genome Browser format description. Results were checked line by line against an independent Python reference implementation.

Data size

Runs in your browser; annotation files up to roughly 100 MB (about 300,000 lines) are recommended. For a complete human GTF (about 1.5 GB uncompressed), extract the chromosomes or genes you need first.