Target Prediction Algorithms and Bioinformatics Resources for miRNA Studies - MicroRNAs as Tools in Biopharmaceutical Production - page 36

algorithm also ranks the outputted prediction based on a “context + score” calculated
from the target type, AU composition, position, site abundance and seed pairing
stability (Garcia et al. 2011 ; Grimson et al. 2007 ).

3.3.5 PITA

The PITA algorithm extends the thermodynamic calculation to consider target ac-
cessibility with respect to transcript secondary structure (Kertesz et al. 2007 ). The
initial stage of PITA predicts targets via seed region complementarity to the mRNA
3 UTR over at least seven bases for human, mouse, worm and fly followed by the
application of a thermodynamic model to calculate G, a comparison of the free
energy gained from binding of the miRNA to the potential target gene G duplex and
the cost of un-pairing the target for miRNA accessibility G open . The user can also
limit results to conserved miRNAs and mRNA if desired.

3.3.6 RNAhybrid

RNAhybrid attempts to locate the most likely target site with the miRNA seed through
the identification of thermodynamically stable matches in the mRNA 3 UTR (Kruger
and Rehmsmeier 2006 ; Rehmsmeier et al. 2004 ). RNAhybrid does not assess pre-
computed alignments to the seed region but rather identifies any regions that have the
ability to form duplexes with a given miRNA. The algorithm also applies a statistical
adjustment to correct for the sequence length of the target and miRNA sequence and
assigns a p-value. In addition, the significance of multiple and conserved sites can
be computed using Poisson statistics.

3.3.7 RNA22

The RNA22 algorithm (Miranda et al. 2006 ) differs from competing methods in
that target prediction does not use evolutionary filtering allowing the identification
of species specific miRNA-mRNA interactions. In addition, RNA22 does not rely
on locating the reverse complement of the miRNA seed region but attempts to find
overrepresented statistical patterns derived from the mature miRNA within the tran-
script of interest potentially identifying unknown miRNA binding sites (RNA22 can
assess the 3 UTR, 5 UTR and CDS). The initial phase of RNA22 identifies targets
sites by identifying motifs within the mature miRNA. The reverse complement of
statistically significant motifs are compared to the transcript to locate potential bind-
ing sites. As the patterns located in stage one of the algorithm contain redundancy,
one or more of these patterns can map to a single transcript. If a pattern aligns at
a particular location a vote is cast and if multiple patterns align to a location (e.g.
votes

≥ 30) the area is identified as a “target island” for potential miRNA binding. To