Markup | Genome Research

Table 1.

Performance Comparison for Wormseq (325 genes, 2254 exons)

Precise Identification of Exons and Complete Gene Structures
Program		Exon-level accuracy		Gene-level accuracy
Program		Sn	Sp	Av.	ME	WE	Sn	Sp	Av.	MG	WG
GAZE_std	0.84	0.77	0.80	12	303	0.35	0.35	0.35	4	25
GAZE_tsplice	0.86	0.80	0.83	12	273	0.47	0.42	0.44	4	34
GAZE_EST	0.90	0.84	0.87	8	222	0.59	0.53	0.56	3	31
Fgenesh	0.88	0.80	0.84	5	319	0.51	0.42	0.47	2	57

Breakdown of Exon-Level Accuracy by Exon Type
Program		Initial		Internal		Terminal		Single
Program		Sn	Sp	Av.	Sn	Sp	Av.	Sn	Sp	Av.	Sn	Sp	Av.
GAZE_std	0.57	0.56	0.57	0.90	0.80	0.85	0.78	0.78	0.78	0.94	0.71	0.83
GAZE_tsplice	0.72	0.67	0.70	0.89	0.83	0.86	0.81	0.74	0.78	0.94	0.59	0.77
GAZE_EST	0.79	0.74	0.77	0.92	0.86	0.89	0.85	0.80	0.83	0.94	0.73	0.84
Fgenesh	0.75	0.61	0.68	0.92	0.87	0.90	0.84	0.68	0.71	0.63	0.77	0.70

[i] Sensitivity (Sn) is the proportion of confirmed genes or exons that are predicted correctly. Specificity (Sp) is the proportion of predicted genes or exons that are correct. Average (Av.) is (Sn + Sp)/2. The values reported are for the exact identification of exons and complete gene structures. Missing exons (ME) are the number of correct exons with no overlap to any predicted exons, and wrong exons (WE) are the number of predicted exons with no overlap to any correct exons. Corresponding values are presented at the gene level (MG, WG).