The Ensembl gtf file contains the comprehensive gene and transcript information for model organisms e.g. human and mouse. It can be used in RNA-Seq alignment and quantification programs such as STAR.

• Downloading the appropriate GTF file
• Formatting of the GTF file
• Features inside the GTF file
  • Extract features of interest from GTF using the command line
  • Import the GTF file into R

Downloading the appropriate GTF file

• The file can be accessed under the ‘Gene sets’ column of the table of the handy Ensembl FTP downloads website.
• for example, you can download the current versions of the mouse and human gtf files with the following commands:

curl ftp://ftp.ensembl.org/pub/release-94/gtf/mus_musculus/Mus_musculus.GRCm38.94.gtf.gz -o Mus_musculus.GRCm38.94.gtf.gz
curl ftp://ftp.ensembl.org/pub/release-94/gtf/homo_sapiens/Homo_sapiens.GRCh38.94.gtf.gz -o Homo_sapiens.GRCh38.94.gtf.gz

Formatting of the GTF file

• the GTF file is a specific type following the GFF annotation format.
• there’s actually a mixture of tabs, colons etc. used to separate fields, which you can see by viewing the file as below:

[me@XXXXX ]$ head Homo_sapiens.GRCh38.94.gtf
#!genome-build GRCh38.p12
#!genome-version GRCh38
#!genome-date 2013-12
#!genome-build-accession NCBI:GCA_000001405.27
#!genebuild-last-updated 2018-07
1	havana	gene	11869	14409	.	+	.	gene_id "ENSG00000223972"; gene_version "5"; gene_name "DDX11L1"; gene_source "havana"; gene_biotype "transcribed_unprocessed_pseudogene";
1	havana	transcript	11869	14409	.	+	.	gene_id "ENSG00000223972"; gene_version "5"; transcript_id "ENST00000456328"; transcript_version "2"; gene_name "DDX11L1"; gene_source "havana"; gene_biotype "transcribed_unprocessed_pseudogene"; transcript_name "DDX11L1-202"; transcript_source "havana"; transcript_biotype "processed_transcript"; tag "basic"; transcript_support_level "1";

Features inside the GTF file

• The third column lists the feature types, which you can also confirm with this small script from Benn:

awk '{print $3}' Homo_sapiens.GRCh38.87.gtf | sort | uniq
CDS
Selenocysteine
exon
five_prime_utr
gene
start_codon
stop_codon
three_prime_utr
transcript

• These seem to be best defined in the gtf submission guidelines under the heading ‘Attributes/Annotation features’. You’ll notice it refers to them as ‘SO’ features which actually corresponds to the Sequence Ontology.
• This is also pointed out on the answer post by Istvan Albert. Indeed searching for terms gives a table containing a brief definition and the children and parents of that term which is a bit more informative. The below image shows for example that UTR is considered a portion of the mature transcript.

Extract features of interest from GTF using the command line

• The Gencode documentation has some beginner short scripts for doing this with awk within the section ‘Examples for fetching specific parts from the file’ here.
• It’s also really nicely outlined in the series of posts by Nacho Caballero here.

Import the GTF file into R

• For downstream applications, the package rtracklayer has some handy import function specifically for gtf files.
• for the given FILE_PATH you can do something like this:

# import gtf file -------------------------------------

mm_gtf <- rtracklayer::import('FILE_PATH/Mus_musculus.GRCm38.94.gtf')
head(mm_gtf)


# extract gene features from the gtf -------------------------------------
unique(levels(mm_gtf$type))
genes_only <- mm_gtf[mm_gtf$type == "gene",]
head(genes_only)