Computational prediction of miRNA precursors and miRNA targets
Abstract
MicroRNA is small 22 nucleotides long non-coding RNA which regulates genes
by targeting mRNA especially the 3’ UTR region. Identification of miRNAs and their
targets by laboratory experiment has had limited success especially for lowly expressed
therefore computational prediction approaches are needed. In this study grouping
technique for miRNA precursor prediction is introduced. Compared with global
alignment, grouping miRNA by classes yield a better sensitivity with very high
specificity for pre-miRNA prediction even when a simple positional based secondary and
primary structure alignment are used.
The program TarSpec was developed to predict miR-1a and miR-124 targets
based on common features of miRNA and target binding characteristics observed from
alignment between miRNA and the 3’UTR targets. TarSpec obtained 78% and 77%
sensitivity for miR-1a and Mir-124 targets, and 98% specificity for both. TarSpec was
used to scan the Platypus 3’UTR regions. This approach predicted 734 novel potential
target of miR-1a in Platypus 3’UTR regions where 98 of them are in the well annotated
chromosomal region and 124 novel potential target of miR-124 where 32 of them are in
the well annotated chromosomal region.
Some miRNAs are derivations of transposable elements (TE). In human these TE
derived miRNAs have a potential to regulate thousands of human genes. Therefore TEs
as potential miRNA targets were investigated using an L2 derived miRNA miR-28. Three
different miRNA target prediction programs miTarget, miRanda, and RNAhybrid were
used to predict a potential miRNA-28 targets in human L2 transposable elements. It was
demonstrated that the human TE is also a potential target for miRNAs; subsequently
1,094 of potential target were predicted in human L2.