Date: 23 Oct 1999 http://www.fruitfly.org/seq_tools/datasets/Human


This directory contains a set of GenBank flat-file format entries of
multiple exon and single exon genes to be used to test and train
gene-finding algorithms for human genomic DNA.

UCSC and LBNL provide these data sets "as is" in an effort to create a
common data set to be used by all gene-finding algorithms.  We
encourage others to compare their results using these data sets.

Accompanying the data set is a ".sets" file listing 7 test/train
subsets.  These subsets can be used for cross-validation.


The files in this directory are:

	multi_exon_GB.dat.gz		GenBank format file 462 multi exon genes
	multi_exon_GB.sets		proposed cross-validation sets

For specific training we have also split the complete Genbank entries 
into various gene feature sets:

	./CDS_v105/			complete CDS
	./intron_v105/			complete introns
	./exon_v105/			complete exons
	./genomic_DNA_v105/		complete genomic DNA from start codon to stop codon
	./coding_data/			all possible unrelated CDSes in Genbank

	single_exon_GB.dat.gz		GenBank format file 331 single exon genes
	single_exon_GB.sets		proposed cross-validation sets


We also provide the Genie data sets from 1995 and 1996 in 
	./GENIE_95
and
	./GENIE_96

The published Bursett/Guigo in GENOMICS '95 can be obtained at 
	ftp://www-hgc.lbl.gov/pub/genesets/OtherDataSets/GUIGO_96


In addition, several Perl and shell routines for generating
cross-validated sets are provided in the scripts directory
(http://www.fruitfly.org/seq_tools/datasets/scripts/).

=====
Martin Reese (LBNL) martinr@bdgp.lbl.gov
With help from:
David Kulp (UCSC) dkulp@cse.ucsc.edu
Andrew Gentles (Stanford)
Uwe Ohler (UCB) ohler@fruitfly.berkeley.edu