Table 1

Results for deletion experiment. The first column is for TWINSCAN 2.03 and the remaining 3 columns are for TWINSCAN_EST. The second column is for the TWINSCAN_EST performance with empty ESTseq, i.e., all bases in ESTseqs are 'N's. For the third and fourth column, 10% of genes in the annotation were set to "N"s. The third column is for TWINSCAN_EST's performance on the 10% of genes with masked ESTseqs and the last column is for the 90% of genes with unmasked ESTseqs. Results show that EST alignments improve the prediction accuracy and do not compromise the capability to predict novel genes where EST alignments do not exist (column 2). Specificities are based on predictions that overlap with annotations by at least 1 bp.

TWINSCAN2.03

TWINSCAN_EST

Blank ESTseq

10% with ESTseq masked

90% with ESTseq unmasked

Gene_sn

60.6

61.3

63.0

74.7

Gene_sp

58.6

59.8

60.5

71.5

Exon_sn

86.9

86.2

86.4

91.5

Exon_sp

79.5

80.8

81.1

87.0

Wei and Brent BMC Bioinformatics 2006 7:327   doi:10.1186/1471-2105-7-327