GFF3 is the successor to the GFF2 feature file format. Its main
advantage is that it uses the Sequence Ontology to describe features
on the genome in a standard and predictable way. For details, see
http://song.sourceforge.net/ and http://song.sourceforge.net/gff3.shtml

The GBrowse Bio::DB::GFF adaptor was developed for GFF2 and its
support for GFF3 is poor. The main issue is that Bio::DB::GFF is only
able to support features that have two levels, for example an mRNA
transcript and its exons, while GFF3 allows features to contain
unlimited levels of features and subfeatures. This feature is most
commonly used to describe genes:

	gene A
	   mRNA A.1
              CDS A.1.1
              CDS A.1.2
              CDS A.1.3
	   mRNA A.2
              CDS A.2.1
              CDS A.2.2
              CDS A.2.3

This data model is encouraged by GFF3 and is used by FlyBase;
unfortunately Bio::DB::GFF cannot store such gene features without
hacking the FlyBase GFF3 files.

Bioperl version 1.5.2 has database support for GFF3 format via the
Bio::DB::SeqFeature::Store module. Currently only MySQL, BerkeleyDB
and in-memory implementations are available. Other implementations
(PostgreSQL, etc) may be written at a future data.

Here are instructions for bringing up the Fly GFF3 annotations.

1) Make sure you are using Bioperl 1.5.2 or higher:
	
   % perl -MBio::Perl -e 'print $Bio::Perl::VERSION'
   1.005002

2) Download the GFF3 files and FASTA files to load and put them in a
directory somewhere.

3) Create a mysql database to hold the data, and set up the
appropriate privileges:

 mysqladmin -uroot -p<passwd> create database flygff3
 mysql -uroot -p<passwd> -e 'grant all privileges on flygff3.* to <mylogin>@localhost'
 mysql -uroot -p<passwd> -e 'grant select on flygff3.* to nobody@localhost'


4) Use the bp_seqfeature_load.pl script to load the GFF3 and FASTA
data:

 cd <directory_containing_the_files>
 bp_seqfeature_load.pl -d flygff3 -c -f *.gff.gz *.fasta

The -d option gives the name of the database to load ("flygff3").
The -c flag tells the script to initialize the database.
The -f flag tells the script to use "fast" loading.

You may see a small number of warnings about features with the same
IDs not occurring in contiguous order. You can safely ignore these
warnings.

5) Copy the 09.fly.gff3.conf config file into your gbrowse.conf
directory. This file can be found in the Generic-Genome-Browser
distribution directory under contrib/conf_files/.

The distributed conf file expects the database to be named "flygff3"
and to require no password to be readable by the "nobody" user. Please
modify it if necessary. For example, to add a username and password:

 db_adaptor    = Bio::DB::SeqFeature::Store
 db_args       = -adaptor DBI::mysql
	         -dsn     dbi:mysql:database=flygff3
                 -user    fred
                 -pass    secretpassword

You should be able to browse the database now!

6) To use the in-memory GFF3 database adaptor do the following:

 % mkdir /var/www/htdocs/gbrowse/databases/volvox_gff3
 % cp docs/tutorial/data_files/volvox.gff3 /var/www/htdocs/gbrowse/databases/volvox_gff3

and edit the volvox.gff3.conf file to read:

 db_adaptor    = Bio::DB::SeqFeature::Store
 db_args       = -adaptor memory
	         -dsn     /var/www/htdocs/gbrowse/databaess/volvox_gff3

Be sure to change the database path to be appropriate for the location
of the gbrowse document directory on your system.

If you use the memory adaptor for large sequences, you will see better
performance if you separate out the DNA part of the GFF3 file from the
annotation part (the volvox.gff3 file has both annotations and
DNA). Put the DNA into one or more .fasta files in the same directory
as the .gff3 file(s). This is the same as the traditional way of
creating a GFF2 in-memory database.

8) For more information:

The 09.fly.gff3.conf file contains brief comments describing the
changes needed to make GBrowse work well with
Bio::DB::SeqFeature::Store, the most important of which is the lack of
aggregators in the latter.

A version of the tutorial files adapted for GFF3 use can be found in
docs/tutorial/data_files/volvox.gff3 and
docs/tutorial/conf_files/volvox.gff3.conf. To load the data, simply:

 % bp_seqfeature_load.pl -d volvoxgff3 -c -f docs/tutorial/data_files/volvox.gff3

and install the volvox.gff3.conf file. This file is heavily annotated
with instructions on how to use the GFF3 database with GBrowse.

Please report all problems to gmod-gbrowse@lists.sourceforge.net.

Good luck!

Lincoln Stein