Target Prediction Algorithms and Bioinformatics Resources for miRNA Studies - MicroRNAs as Tools in Biopharmaceutical Production - page 30

cross-species microarrays (Gammell et al. 2007 ), qPCR and next generation se-
quencing (NGS) platforms (Hackl et al. 2011 ; Hammond et al. 2011 ; Johnson et al.
2011 ) to study these small RNAs in CHO. To date, miRNAs have been implicated in
the control of several industrially relevant phenotypes including growth (Bort et al.
2011 ; Gammell et al. 2007 ), apoptosis (Druz et al. 2011 ), and recombinant protein
production (Barron et al. 2011 ; Hammond et al. 2011 ; Lin et al. 2010 ). As we enter
the next phase of CHO miRNA research computational tools will become increas-
ingly important in unravelling the effects of miRNA expression on the transcriptome,
proteome and ultimately bioprocess phenotypes.
Bioinformatics plays a crucial role in the study of miRNA biology. Researchers
utilise a variety of supporting computational methods in a range of areas from se-
quence analysis to detecting differential expression of miRNAs. While pre-existing
tools were readily available for some of these tasks, a new class of in-silico methods
to predict miRNA interactions with mRNA (therefore predicting miRNA function)
was required. The development of these algorithms was driven by the breadth and
complexity of animal miRNA dependant post-transcriptional control ( 30%ofall

protein coding genes are estimated to be targeted by one or more miRNA (Filipowicz
et al. 2008 )) along with the extensive wet-lab experimentation required to confirm a
miRNA target.
The aim of this chapter is to describe miRNA target prediction algorithms; we be-
gin by describing the currently accepted rules of miRNA target interaction (Sect. 3.2),
followed by a detailed treatment of the most popular algorithms (Sect. 3.3). Publicly
available data repositories containing precompiled lists of predicted targets, validated
targets and sequence information are described (Sect. 3.4); the chapter closes with
a discussion of algorithm performance (Sect. 3.5) and examples of their application
in CHO studies (Sect. 3.6).

∼

3.2 Principles of MicroRNA Target Prediction

The prediction of miRNA targets in plants is generally straightforward. Here
miRNAs tend to bind their mRNA targets with near perfect complementarity induc-
ing cleavage of the transcript (Rhoades et al. 2002 ), and while near-perfect matches
between metazoan miRNAs and their respective targets can be found, it is unusual.
Prediction of target interactions in animals is generally more challenging due to
the intricacy of target recognition where sequences often contain gaps, mismatches
and G:U base pairs in multiple positions (Bartel 2009 ). The complexity of animal
miRNA regulation is reflected by estimates that a single miRNA targets an average
of 100-200 mRNAs (Krek et al. 2005 ), a single mRNA transcript can be targeted
by hundreds of miRNAs (Friedman et al. 2009 ) and multiple miRNAs can cooper-
atively repress a range of targets (Wu et al. 2010 ). Experimental methodologies to
confirm direct miRNA regulation have advanced over recent years; however the num-
ber of confirmed interactions with mRNA remains relatively small. Computational
target prediction remains a critical aspect of assessing global miRNA function and