Automatic generation of gene finders for eukaryotic species

BMC Bioinformatics

Table 1 Performance evaluation. NSN: nucleotide sensitivity, NSP: nucleotide specificity, ESN: exon sensitivity, ESP: exon specificity, ME: missed exons, WE: wrong exons. NSN is defined as the percentage of annotated coding bases predicted as coding and NSP as the percentage of predicted coding bases annotated as as coding. ESN and ESP reflect analogously how well the methods predict exons exactly right. Superscripts on the species names indicate the type of generated gene model. 1: Full model. 2: 3' UTR exons and introns not modelled. 3: No UTR exons and introns modelled. 4: UTR exons and introns as well as UTR part of first and last exons not modelled. Subscripts indicate whether shared length distributions are used. 1: No shared distributions. 2: Internal and single coding exons share distribution. 3: all coding coding exons share distribution

Species	Predictor	NSN	NSP	ESN	ESP	ME	WE
$C . e l e g a n s_{2}^{3}$	Agene	95	92	86	85	5	9
	GeneID	95	86	75	68	6	16
$D . m e l a n o g a s t e r_{1}^{1}$	Agene	87	87	64	68	13	15
	Augustus	85	91	67	72	15	11
$A . t h a l i a n a_{1}^{1}$	Agene	85	89	67	74	14	11
	Genscan	82	83	49	51	17	18
$N . c r a s s a_{1}^{4}$	Agene	76	95	57	76	25	7
	GeneID	24	95	14	83	76	8
$A . n i d u l a n s_{2}^{4}$	Agene	84	93	59	70	19	9
	GeneID	88	86	49	50	18	19
$H . s a p i e n s_{1}^{1}$	Agene	63	56	47	40	38	48
	Genscan	87	60	63	47	14	39
$M . m u s c u l u s_{1}^{2}$	Agene	73	88	56	68	28	14
	Genscan	88	82	69	68	14	17
$D . r e r i o_{2}^{3}$	Agene	88	89	67	70	13	13
	Genscan	91	87	67	69	12	10
$A . g a m b i a e_{1}^{3}$	Agene	81	83	52	63	18	15
$D . h a n s e n i i_{2}^{4}$	Agene	82	85	64	68	17	19
$K . l a c t i s_{3}^{4}$	Agene	78	91	69	79	21	11
$C . g l a b r a t a_{3}^{4}$	Agene	91	94	82	82	9	9
$U . m a y d i s_{1}^{4}$	Agene	87	94	61	69	14	9

ISSN: 1471-2105