GENSCAN gene predictions parsed out of following XML file:
ftp://ftp.fruitfly.org/pub/download/dmel_RELEASE3-1/Annotations_and_Evidence/whole_genome_annots+evidence_dmel_RELEASE3-1.GAME-XML.tar.gz

to yield genscan_predictions.gff.

Gene predictions with even a single base pair overlap were merged into a single prediction to yield genscan_preds_genemerged.gff.

This merged file was used in all subsequent analysis.  Note that the lenient merging criteria implies a conservative estimate for the number of novel genes confirmed by our expressed nonexon probes (NEPs).



SUMMARY OF FILES:
genscan_predictions.gff - see above.

genscan_preds_genemerged.gff - see above.

novel_exons_confirmed_by_NEPs.gff - list of 1155 exons absent from the GADFLY 3.1 annotation that are predicted by GENSCAN and confirmed by at least one overlapping, significantly expressed NEP (non-exon probe).  The 1155 include 369 novel exons belonging to annotated genes, and another 786 exons belong to completely novel predicted genes.

novel_exons_of_known_genes_confirmed_by_NEPs.gff - the 369 exons belonging to annotated genes.

novel_genes_confirmed_by_NEPs.gff - GENSCAN-predicted genes absent from GADFLY and confirmed by at least one NEP (the 786 exons mentioned above are a subset of this list).

novel_upstreamsegments_confirmed_by_NEPs - Regions where GENSCAN predicts a longer 5' boundary for an exon relative to the 3.1 annotation.  The 4 columns are chromosome, start position, end position, and strand designating this region.  The corresponding annotated exon would start at (end position + 1).

novel_downstreamsegments_confirmed_by_NEPs - Regions where GENSCAN predicts a longer 3' boundary for an exon relative to the 3.1 annotation.  The 4 columns are chromosome, start position, end position, and strand designating this region.  The corresponding annotated exon would end at (start position - 1).