Computational discovery of animal small RNA genes and targets

Though recently discovered, small RNAs appear to play a wealth of regulatory
roles, being involved in degradation of target mRNAs, translation silencing of
target genes, chromatin remodeling and transposon silencing. Presented here
are the computational tools that I developed to annotate and characterize
small RNA genes and to identify their targets. One of these tools is oligomap,
a novel software for fast and exhaustive identi�cation of nearly-perfect matches
of small RNAs in sequence databases. Oligomap is part of an automated annotation
pipeline used in our laboratory to annotate small RNA sequences. The
application of these tools to samples of small RNAs obtained from mouse and
human germ cells together with subsequent computational analyses lead to
the discovery of a new class of small RNAs which are now called piRNAs. The
computational analysis revealed that piRNAs have a strong uridine preference
at their 5' end, that unlike miRNAs, piRNAs are not excised from fold-back
precursors but rather from long primary transcripts, and that the genome organization of their genes is conserved between human and mouse even though
piRNAs on the sequence level are poorly conserved. In vertebrates, the most
studied class of small regulatory RNAs are the miRNAs which bind to mRNAs
and block translation. A computational framework is introduced to identify
miRNA targets in mammals,
ies, worms and �sh. The method uses extensive
cross species conservation information to predict miRNA binding sites that
are under evolutionary pressure. A downstream analysis of predicted miRNA
targets revealed novel properties of miRNA target sites, one of which is a
positional bias of miRNA target sites in long mammalian 3' untranslated regions.
Intersection of our predictions with biochemical pathway annotation
data suggested novel functions for some of the miRNAs. To gain further insights
into the mechanism of miRNA targeting, I studied microarray data
obtained in siRNA experiments. SiRNAs have been shown to produce o�-
targets that resemble miRNA targets. This analysis suggests the presence of
additional determinants of miRNA target site functionality (beyond complementarity
between the miRNA 5' end and the target) in the close vicinity
(about 150 nucleotides) of the miRNA-complementary site. Finally, as part
of a study aiming to reduce siRNA o�-target e�ects by introducing chemical
modi�cations in the siRNA, I performed microarray data analysis of siRNA
transfection experiments. Presented are the methods used to quantify o�-
target activity of siRNAs carrying di�erent types of chemical modi�cations.
The analysis revealed that o�-targets caused by the passenger strand of the
siRNA can be reduced by 5'-O-methylation.