In silico miRNA prediction in metazoan genomes: balancing between sensitivity and specificity

BMC Genomics

Table 1 Subset of 18 most informative miRNA hairpin descriptors

Descriptor	Explanation	Bound. ^a	Type^b	Discriminative power^c	K^d
bulgeRatio	ratio asymmetrical bulges vs. stem length	↑	str	1.45	0.416
dP	adjusted base pairing propensity (dP)	↓	str	2.28	0.417
largest bulge	longest bulge in stem (nt)	↑	str	1.74	0.343
longest match-stretch	longest match-stretch in stem (nt)	↓	str	1.20	0.336
Looplength	central loop length (nt)	↑	str	1.17 (u)	0.191
max match count	matches in 24 nt	↓	str	2.75	0.477
MFEahl index [39]	MFEahl corrected for GC- Content	↓	str	4.75	0.706
Q [37, 39]	Normalized Shannon entropy (Q)	↑	str	3.01	0.844
stem length	stem length	↑↓	str	1.29 (l)	0.404
GAsurplusCU	surplus of GA over CU in sequence	↑↓	seq	1.12 (u) 1.05 (l)	0.195
GsurplusC	surplus of G over C in sequence	↓	seq	1.12	0.995
polyA	longest poly-A stretch (nt)	↑	seq	1.58	0.834
polyNucHairpin	longest mono-nucleotide stretch (nt) in the hairpin	↑	seq	1.64	0.846
polyU	longest poly-U stretch (nt)	↑	seq	1.53	0.540
SCS-di	Di-nucleotide Sequence Complexity (-)	↑	seq	1.77	0.557
SCS-mono	Mono-nucleotide Sequence Complexity (-)	↓	seq	1.60	0.317
GU-match contribution	ratio of GU-matches vs. all matches	↑	mix	1.28	0.173
MFEahl (dG) [37, 39]	MFE Adjusted for hairpin length	↓	mix	13.33	0.742

A detailed explanation of all 40 descriptors is given in the additional information [see Additional files 1 and 2].
^a Boundary; indication of extreme tail of descriptor distribution that was transformed into S < 1 fraction. Symbols denote: ↓ lower tail; ↑ upper tail, and; ↓↑ both tails.
^b Type; descriptor based on structural (str), sequence (seq) or both structural and sequence (mix) properties of the hairpin.
^c Discriminative power; expressed at 95% sensitivity, measured on the taxonomic set Metazoa (3,902 miRNA hairpins, positives) and genomic hairpins in C. elegans (3,526,115 hairpins, negatives). * Discriminative power of descriptor 'Z' was measured on 25,599 identified hairpins in four viruses.
^d K; highest Cohen's kappa coefficient [41] with another descriptor, measured on the taxonomic set Metazoa with S < 1 cut-off of 95%.

ISSN: 1471-2164