Genome-wide identification of specific oligonucleotides using artificial neural network and computational genomic analysis

Table 2 Performance comparison with and without an artificial neural network.

Oligo length	Procedure	Sensitivity factor	Success rate (%)^a	Cross homology ^b	Execution time (hours)
70-mer	IAB ^c	0.2	94	0.56	0.50
		0.3	95	0.56	0.71
		0.4	95	0.56	0.90
	BLAST with U_d^d		95	0.56	2.13
	Pure BLAST		95	0.61	3.69
50-mer	IAB	0.2	91	0.64	0.20
		0.3	93	0.64	0.31
		0.4	93	0.64	0.32
	BLAST with U_d		93	0.64	0.70
	Pure BLAST		93	0.69	2.19
25-mer W = 11^e	IAB	0.2	94	0.79	0.19
		0.3	94	0.79	0.26
		0.4	94	0.79	0.33
	BLAST with U_d		94	0.79	0.78
	Pure BLAST		94	0.80	1.62
25-mer W = 8	IAB	0.2	93	0.80	0.36
		0.3	93	0.80	0.50
		0.4	93	0.80	0.64
	BLAST with U_d		93	0.80	1.52
	Pure BLAST		93	0.81	3.51
25-mer W = 5	IAB	0.2	92	0.81	1.00
		0.3	92	0.81	1.39
		0.4	92	0.81	1.79
	BLAST with U_d		92	0.81	4.25
	Pure BLAST		92	0.81	9.69

^aThe success rate is the percentage of tentative human consensus sequences on which the procedure can find the specific oligo (where the cross homology is less than the threshold). ^bThe cross homology of a specific oligo is determined by the similarity between the specific oligo and its best homology in the non-target sequence. ^cIAB, integration of artificial neural network (ANN) and basic local alignment search tool (BLAST).^dBLAST with U_d, BLAST search with the density of unique subsequences. ^eW is word length in the parameters of BLAST.

ISSN: 1471-2105