GFF3

GFF3 is a text-based file format for representing genomic annotations. The major difference from BED is that GFF3 is more structured and can include sequences in the FASTA file format.

The GFF3 package supports I/O for GFF3 by providing the following three types:

• Reader type: GFF3.Reader
• Writer type: GFF3.Writer
• Element type: GFF3.Record

A GFF3 file may contain directives and/or comments in addition to genomic features. These lines are skipped by default but you can control the behavior by passing keyword arguments to GFF3.Reader. See the docstring for details.

Examples

Here is a common workflow to iterate over all records in a GFF3 file:

# Import the GFF3 module.
using GFF3

# Open a GFF3 file.
reader = open(GFF3.Reader, "data.gff3")

# Iterate over records.
for record in reader
    # Do something on record (see Accessors section).
    seqid = GFF3.seqid(record)
    # ...
end

# Finally, close the reader
close(reader)

The iterator interface demonstrated above allocates an object for each record and that may be a bottleneck of reading data from a file. In-place reading reuses a pre-allocated object for every record and less memory allocation happens in reading:

# Import the GFF3 module.
using GFF3

# Open a GFF3 file.
reader = open(GFF3.Reader, "data.gff3")

# Pre-allocate record.
record = GFF3.Record()

# Iterate over records.
while !eof(reader)
    empty!(record)
    read!(reader, record)
    # do something
end

# Finally, close the reader.
close(reader)

If you are interested in directives (which starts with '#') in addition to genomic features, you need to pass skip_directives=false when initializing a GFF3 constructor:

# Set skip_directives to true (this is set to false by default).
reader = GFF3.Reader(open("data.gff3"), skip_directives=false)
for record in record
    # Branch by record type.
    if GFF3.isfeature(record)
        # ...
    elseif GFF3.isdirective(record)
        # ...
    else
        # This never happens.
        assert(false)
    end
end
close(reader)

GFF3.jl supports tabix to retrieve records overlapping with a specific interval. First you need to create a block compression file from a GFF3 file using bgzip and then index it using tabix.

cat data.gff3 | grep -v "^#" | sort -k1,1 -k4,4n | bgzip >data.gff3.bgz
tabix data.gff3.bgz  # this creates data.gff3.bgz.tbi

Then you can read the block compression file as follows:

# ...

# Read the block compression gzip file.
reader = GFF3.Reader("data.gff3.bgz")
for record in eachoverlap(reader, Interval("chr1", 250_000, 300_000))
    # Each record overlap the query interval.
    # ...
end