PyGFF - Useful tools for working with GTF/GFF3 format files.

How to install gfftools

Install PyGFF from git repo via pip:

pip install git+https://github.com/altairwei/PyGFF.git

How to use gfftools

gfftools contains a series of subcommands, includes stats, conv, filter and seq.

Filter GFF features with given conditions

Match specified fields

We can specify multiple fields of the same type, and as long as the feature meets one of them, it will be preserved.

For example:

gfftools filter --seqid 1 --seqid 3 --source ensembl --type CDS Homo_sapiens.GRCh38.99.gtf > chr1_or_chr3_CDS_from_ensembl.gtf

All attributes need to be matched

We can specify multiple attributes at the same time, and only features that satisfy all of them will be preserved.

For example:

gfftools filter --attributes gene_version=5 --attributes gene_biotype=lncRNA > lncRNA.gtf

Keep features in a given region

Regions can be specified as: SEQID[:STARTPOS[-ENDPOS]] and all position coordinates are 1-based.

Important note: Only records whose start and end points are both contained within this region will be kept.

Examples of region specifications:

• chr1 - Output all features belongs to the reference sequence named "chr1".
• chr2:1000000 - The region on "chr2" beginning at base position 1,000,000 and ending at the end of the chromosome.
• chr2:-2000 - The region on "chr2" beginning at the start of the chromosome and ending at base position 2,000.
• chr3:1000-2000 - The 1001bp region on "chr3" beginning at base position 1,000 and ending at base position 2,000.

For example:

gfftools filter --region chr2:1000000 --region chr3:1000-2000 Homo_sapiens.GRCh38.99.gtf > chr2_chr3_features.gtf

Evaluate python condition expression

The filter will execute a user-specified python conditional expression for each feature, and the feature will be preserved if the expression is true. The environment in which the expression is executed contains 9 predefined variables.

seqid: str # chromosome name
source: str # Where the feature is generated
type: str # Feature type, CDS, exon and so on
start: int # Start position, 1-based indexing
end: int # End position, 1-based indexing
score: str # The score of the feature
strand: str # strand of DNA
phase: str # aka. frame
attributes: dict # Feature attributes are wraped in a dict.

For example:

gfftools filter -e "(end - start + 1) % 3 == 0" Homo_sapiens.GRCh38.99.gtf > triad.gtf

Extract gene sequences from the genome based on GFF file

In general, we want to extract gene sequences from genome with a small number of features, so the gfftools seq command supports the same GFF filter as gfftools filter.

gfftools seq -a gene_id=ENSG00000177133 -g Homo_sapiens.GRCh38.dna.primary_assembly.fa Homo_sapiens.GRCh38.99.gtf > ENSG00000177133.fa

The file ENSG00000177133.fa in above example would be a multi-fasta file with the sequences corresponding to features of filtered GFF content. The sequence names of genome fasta file (in this case, it means Homo_sapiens.GRCh38.dna.primary_assembly.fa) should match what in 1st column of the input GFF file.

There are some flags to control the output format of FASTA file. For example, the following command was used to extract gene sequences of wheat with gene ID as FASTA header.

gfftools seq \
    --type gene \
    -g data/Triticum_aestivum.IWGSC.dna.toplevel.fa \
    --line-length 60 \
    --fasta-header "attributes['ID']" \
    data/Triticum_aestivum.IWGSC.48.gff3 > data/genes.fa

Convert GFF3 to GTF

Work in progress.

Count features' properties of GFF file

gfftools stats Homo_sapiens.GRCh38.99.gtf