|
Genome-wide assessment of ab initio gene prediction methods |
|||||
| Genscan |
Grail |
Fgenes |
Transcript- |
Protein- |
Unconfirmed |
| confirmed |
supported |
exons |
|||
| exons |
exons |
||||
|
|
|||||
| • |
25,619 |
2,890 |
45,025 |
||
| • |
52,644 |
14,685 |
434,409 |
||
| • |
7,791 |
796 |
257,676 |
||
| • |
• |
17,841 |
3,761 |
28,556 |
|
| • |
13,915 |
1,711 |
11,628 |
||
| • |
• |
3,990 |
450 |
49,420 |
|
| • |
• |
• |
53,566 |
9,871 |
20,569 |
| Total exons |
175,366 |
34,164 |
847,283 |
||
|
Three gene prediction programs, Genscan [93], Fgenes [94] and Grail1.3 [95] were used to screen individual genomic contigs. Exons consistently predicted by more than one program are merged into a unique exon index, which is then compared to transcript- and protein-based exons in complete ORFs. Transcript-confirmed exons, overlapping of predicted exons with transcript-based exons; protein-supported exons, predicted exons have at least strong protein homology (E < 10-15); unconfirmed exons, predicted exons have no overlap with transcripts nor protein homology. | |||||
Wright et al. Genome Biology 2001 2:research0025.1 doi:10.1186/gb-2001-2-7-research0025 |
|||||