Biology Reference
In-Depth Information
are most important for miRNA function,
102-104
Xie
et al
.
101
identified
8mers that were strongly conserved in the 3
UTRs of protein-coding
transcripts, and searched for regions in the genome that were reverse
complementary to the 3
′
UTR sequences and could form stable stem-
loops. This procedure generated a set of 129 novel predicted
miRNAs.
Still more miRNA prediction programs are available, particularly
those based on SVMs. Of these, some of the most distinct ones are the
triplet-SVM classifier,
105
which has as features the frequencies of triplets
representing the secondary structure of three contiguous nucleotides;
RNAmicro,
106
which uses a small number of descriptors downstream of a
conservation filter, among them a structure conservation index; and the
Microprocessor SVM,
107
which includes a number of features aiming to
capture the specificity of Drosha processing.
′
7. MicroRNA Target Prediction
7.1. What Does a miRNA Target Site Look Like?
In contrast to lin-4, the founder of the miRNA class, which was discov-
ered by forward genetics and for which the phenotype was known at the
time when the gene was discovered, the vast majority of miRNAs that
have been identified experimentally or predicted computationally do not
yet have an associated function. This generated great interest in compu-
tational predictions of miRNA targets and, not surprisingly, several meth-
ods were already proposed very shortly after miRNA gene prediction
methods were published. Perhaps similar to the situation of miRNA
genes, the main roadblock is that it is still unclear what functional
miRNA target sites look like.
7.2. The miRNA Seed Region
One of the most predictive features that has so far been described for
miRNA target sites is the perfect complementarity with the first seven to
eight nucleotides at the 5
end of the miRNA, a region which has been
called the “nucleus”
108
or “seed”.
81
Lai
109
initially observed that sequence
′
