Table 2

Comparing the enhancement on putative exons with existing models results based on human-mouse sequence conservation.

clearly orthologous exons (TP)

potential non-exons (FP)


size

76,229

1, 518, 082

GENSCAN (w/mouse)

53,217 (69.8%)

54,360 (3.58%)

GENSCAN (w/mouse) S > -0.33

52,682 (69.1%)

14,604 (0.95%)

TWINSCAN (w/mouse)

54,879 (72.0%)

12,276 (0.8%)

TWINSCAN (w/mouse) S > -0.12

54,331 (71.3%)

7,876 (0.5%)

shortHMM S > 0.69 (w/mouse)

74.5%

0.77%


The number of clearly orthologous exons and potential non-exons in the test set are listed in the row of size. The rows of GENSCAN and TWINSCAN list the numbers of putative exons provided by GENSCAN and TWINSCAN respectively. The thresholds for GENSCAN and TWINSCAN are set so that 99% of the correct predictions of GENSCAN and TWINSCAN that have alignments are kept. The percentages in the parentheses are the true positive and false positive rates relative to the sizes of the test sets. The row of shortHMM is cited from [15].

Wu BMC Genomics 2008 9(Suppl 2):S13   doi:10.1186/1471-2164-9-S2-S13

Open Data