Gene prediction in eukaryotes with a generalized hidden Markov model that uses hints from external sources

BMC Bioinformatics

Table 1 Accuracy results on sag178

	sag178	base		exon		gene
		sn	sp	sn	sp	sn	sp
	AUGUSTUS	0.93	0.83	0.79	0.73	0.42	0.38
	GENSCAN	0.94	0.64	0.68	0.45	0.18	0.14
	GENEID	0.89	0.78	0.67	0.60	0.17	0.17
	HMMGene	0.87	0.49	0.71	0.30	0.20	0.07
AUGUSTUS+	EST	0.95	0.85	0.85	0.76	0.49	0.46
	Protein	0.98	0.90	0.91	0.87	0.71	0.68
	Combined	0.95	0.89	0.89	0.85	0.68	0.65
	all hints	0.98	0.93	0.94	0.90	0.82	0.79
	GenomeScan*	0.83	0.77	0.69	0.64	0.37	0.38
	TWINSCAN	0.88	0.82	0.71	0.68	0.20	0.25

Accuracy results on human data set sag178 with 43 sequences and 178 genes, sn = sensitivity = TP/AP, and sp = specificity = TP/PP where TP, AP and PP are the number of true positives, actual positives and predicted positives, 'exon' and 'gene' here refer to the protein-coding parts of exons or genes, respectively. * For the accuracy values of GenomeScan the 7 sequences which are longer than 230 Kb were deleted from the test set. They were too long for the GenomeScan web server.

ISSN: 1471-2105