Small RNA molecules have emerged as important regulators of gene
expression. The most studied among them are the microRNAs (miRNAs),
which are conserved over large evolutionary distances, i.e. from worm
to human, and are involved in many developmental and physiological
processes, such as cell lineage decisions and proliferation,
apoptosis, morphogenesis, fat metabolism, and hormone
secretion. Whereas over 300 miRNA genes have now been identified in
human, experimental identification of their molecular targets has
lagged markedly behind with only a few targets for a few specific
miRNAs currently known. Based on these and a few landmark mutational
studies, a number of researchers have proposed computational methods
for identifying the mRNA targets of miRNAs. The main conclusion that
emerged out of these studies is that functional miRNA binding sites
can be most reliably identified in mRNAs by searching for matches to
the 5' end of the miRNA (``seed'' sequence) that are conserved across
different species. However, the computational approaches have raised
many more questions that remain to be answered. It is, for instance,
unclear what fraction of the large number of mRNAs that contain one or
more matches to a miRNA ``seed'' represents functional targets, how
much the regulatory effect varies across target sites, how the
regulatory effects of small RNAs depend on the concentrations of both
the small RNAs and their targets, and how much the target sets of
related miRNAs overlap (within and across species).

To address these questions we propose to develop new computational
models of miRNA-target interactions and to use these to reconstruct
miRNA-dependent regulatory networks. The basic premise of our
computational approach is that factors beyond the ``seed'' of the
miRNA contribute to the specificity of miRNA action, and that we can
improve the quality of target prediction by incorporating these
additional constraints as well as through more refined models of miRNA
target site evolution. These tools will then allow us to then study
the structure of small RNA-dependent regulatory networks, that include
positive and negative feedback loops about which virtually nothing is
currently known.

This is a particularly opportune time for engaging in such a project
because new high-throughput techniques for detecting miRNA targets in
vivo have become available, large scale expression data for miRNAs and
mRNAs across different tissues are being generated, and new genetic
techniques and reporter essays have been developed to study specific
miRNA-target interactions in detail. By working in close collaboration
with different experimental groups that have pioneered some of the
techniques just mentioned, we will able to directly test our
predictions, and to iteratively use the results to improve our
models. The projects that we propose to carry out are:

1.  Computational models of miRNA-target interaction. We propose to
develop new and more sophisticated models for computationally
predicting the interactions of miRNAs with their targets. The main
improvements of our proposal over existing methods are: (1)
incorporation of explicit models of the evolution of miRNAs and their
targets across related species, taking into account the turnover of
target sites, the phylogenetic relationships between species, and the
nucleotide bias specific to each species; (2) incorporation of
modulator effects induced by the 3' end of the miRNA and protein
co-factors on miRNA function, and (3) explicit modeling of multiple
co-expressed miRNAs and protein co-factors binding to the same mRNA.

2. Characterization of let-7-dependent regulatory network in
C.elegans. Previous efforts to elucidate the roles of miRNAs have
typically looked at individual miRNA-target interactions in
isolation. Studies in C. elegans suggest, however, a complex interplay
between miRNAs, their targets, and the factors regulating miRNA
expression and activity. Here we propose to use an interdisciplinary
approach in which we will iterate computational prediction,
experimental testing and data analysis to obtain a comprehensive view
of the interaction network of the C. elegans let-7 family of
miRNAs. This will include the miRNAs themselves, the regulated targets
and the co-factors that modulate the miRNA activity.

3. Functional characterization of piRNAs.  piRNAs represent a novel
class of small regulatory RNAs presumed to regulate male meiosis that
were discovered in a collaboration between our group and the
laboratory of our collaborator Tom Tuschl at the Rockefeller
University. Here we propose to take advantage of the genome
organization of ctRNA genes and of the experience that we will
accumulate in predicting miRNA target sites to study the role of the
ctRNA-dependent regulatory network in germ cell differentiation.

Through these activities we believe that we can make significant
contributions to the understanding of the function of small RNAs in
cell differentiation and development.