Gene models

What it drawsCopy Anchor Link

A transcript annotation file as the transcripts it describes.

A GTF line states one interval: an exon, a piece of coding sequence, an untranslated region, a start or stop codon. It never states a transcript. The transcript is what those lines add up to: which exons belong together, where the coding sequence begins and ends, and which exons are never translated. This canvas assembles the lines into transcripts and draws them.

• The reading says what the file holds: how many transcripts, of how many genes, on how many sequences, and how many are coding. Below it are only the caveats this file earns (see What this computes), and then how the transcripts were assembled.
• The gene lane draws every transcript on one sequence at a time, in the convention genome browsers use: a thin line for the transcript, short boxes for untranslated exons, tall boxes for coding exons, and chevrons pointing the way the gene is read.
• The transcript table lists every transcript in the file, with its gene, locus, strand, exon count, coding length and whether that length is a whole number of codons. Click a row and the lane draws that transcript alone, with each exon listed: its locus, its length, how much of it is coding, and whether the file stated it or it was built from the coding lines.

What opens in itCopy Anchor Link

GTF (Gene Transfer Format), the nine-column annotation format Ensembl and GENCODE publish. Ensembl's own description: "identical to GFF version 2". Lines are grouped into transcripts by their transcript_id and into genes by their gene_id. A GTF carries no sequence and states no genome assembly, so the canvas asserts neither. Its coordinates are the file's own, shown 1-based.

A GFF3 file is not drawn here. It can carry its own sequence, so it opens as a sequence.

What this computesCopy Anchor Link

• Transcripts. Each transcript_id becomes one transcript spanning its exons. A file does not have to state a transcript line; when it does not, the transcript spans the exons that name it, and the reading says how many transcripts were built this way.
• Exons built from coding lines. Where a file gives CDS and UTR lines but no exon line covering them, the exon is taken as the run those lines cover. The reading counts these, and the exon list marks each one.
• Coding span. From the first base of any CDS or start codon to the last base of any CDS or stop codon, so the stop codon is inside it. This is the convention of a genePred track, where the coding region includes the stop. A transcript with no coding line is non-coding.
• Coding length and whole codons. The coding bases summed over exons. A length that is not a multiple of three is flagged, because it means the annotated CDS is incomplete at one end.
• Lines that belong to no transcript (gene lines, intergenic records, lines with an empty transcript_id) are counted and named in the reading, and are not drawn.

The assembly is checked against UCSC's gtfToGenePred, the converter behind UCSC's own gene tracks: transcript span, coding span and every exon agree for every transcript in the example files.

What it does not doCopy Anchor Link

It does not fetch the genome sequence under a transcript, translate it, or place it on an assembly. To put a GTF's features on a sequence you already have, open that sequence and use Apply features from a file.