Table 2

Sensitivity and specificity of each method at detecting the 1,556 consensus transcripts

Track

Number of models with introns

Consensual models (of 1,556 total)

Sensitivity

Specificity


*AceView

1,792

1,302

84%

73%

Gencode

1,691

1,255

81%

74%

*ECgene

3,851

1,198

77%

31%

AceView

1,630

1,165

75%

71%

*ExonWalk

892

511

33%

57%

*Known Gene

501

432

28%

86%

Exogean

554

404

26%

73%

*RefSeq

342

332

21%

97%

Pairagon

471

310

20%

66%

*Ensembl

427

295

19%

69%

*MGC

323

217

14%

67%

Fgenesh

462

217

14%

47%

*P CCDS

201

152

10%

76%

UP Jigsaw

259

150

10%

58%

*U NscanEst

282

104

7%

37%

UP Augustus

312

100

6%

32%

P Twinscan

547

77

5%

14%

UP GeneMark

551

50

3%

9%

UP SGP2

552

48

3%

9%

UP GeneZilla

477

47

3%

10%

UP ExonHunter

807

41

3%

5%

UP GeneID

267

38

2%

14%

*UP GenScan

395

37

2%

9%

UP Dogfish

204

33

2%

16%

UP Saga

331

18

1%

5%


Sensitivity and specificity of each method at detecting the 1,556 consensus transcripts across the pool of the following evidence based tracks: RefSeq, Known Gene, Ensembl, Gencode, AceView, ECgene and ExonWalk, as in Figure 3a. Coded information has been added in front of the track name: asterisks distinguish standard gene tracks, available genome-wide, from an ENCODE only track; a U track predicts a unique model per gene; P predicts protein coding regions only.

Thierry-Mieg and Thierry-Mieg Genome Biology 2006 7(Suppl 1):S12   doi:10.1186/gb-2006-7-s1-s12

Open Data