Detection of non-coding RNAs on the basis of predicted secondary structure formation free energy change

BMC Bioinformatics

Table 7 Comparison of three ncRNA detection programs on a whole genome screen using WuBLASTn genomic windows.

	probability cutoff for ncRNA classification
	P > 0.5		P > 0.9		P > 0.99
	Dynalign	RNAz	Dynalign	RNAz	Dynalign	RNAz	QRNA
Known ncRNAs found (percent of total known ncRNAs in parentheses)
E. coli (156 ncRNAs known)	147 (94.23)	148 (94.87)	141 (90.38)	140 (89.74)	128 (82.05)	124 (79.49)	121 (77.56)
S. typhi (110 ncRNAs known)	109 (99.09)	108 (98.18)	107 (97.27)	106 (96.36)	103 (93.64)	95 (86.36)	100 (90.91)
Number of contiguous, non-overlapping hits that are not known ncRNAs (i.e. novel ncRNA candidates)
E. coli	2,569	1,898	1,828	1,568	1,211	1,257	1,403
S. typhi	3,936	2,503	2,440	1,986	1,520	1,520	1,611
Number of nucleotides classified as ncRNA that are not in known ncRNAs (i.e. nucleotides in novel ncRNA candidates)
E. coli (each strand = 4,639,675 nt)	324,033	275,704	235,227	221,802	167,157	174,682	206,840
S. typhi (each strand = 4,809,037 nt)	514,650	387,099	352,039	296,608	235,018	220,485	248,724
Total number of nucleotides classified as ncRNA (i.e. nucleotides in both known and unknown ncRNAs)
E. coli (each strand = 4,639,675 nt)	385,611	332,000	292,940	270,267	220,352	209,807	242,534
S. typhi (each strand = 4,809,037 nt)	570,994	436,137	405,022	338,254	283,999	250,610	283,508

QRNA, RNAz, and the Dynalign/LIBSVM classifier are compared in their ability to detect known ncRNA in the E. coli and S. typhi genomes, based on genomic scanning windows prepared using WuBLASTn. For RNAz and the Dynalign/LIBSVM classifier, results are listed for three P value classification cutoffs. "Number of nucleotides" = number of nucleotides on the plus strand + number of nucleotides on the minus strand, not accounting for overlap of complementary strands.

ISSN: 1471-2105