Gene regulatory networks (GRNs) consist of thousands of genes and proteins
which are dynamically interacting with each other. Researchers have investigated
how to uncover these unknown interactions by observing expressions of biological
molecules with various statistical/mathematical methods. Once these regulatory
structures are revealed, it is necessary to understand their dynamical behaviors
since pathway activities could be changed by their given conditions. Therefore,
both the regulatory structure estimation and dynamics modeling of GRNs are essential for biological research.
Generally, GRN dynamics are usually investigated via stochastic models since
molecular interactions are basically discrete and stochastic processes. However,
this stochastic nature requires heavy simulation time to find the steady-state solution of the GRNs where thousands of genes are involved. This large number of
genes also causes difficulties such as dimensionality problem in estimating their
regulatory structure.
This thesis mainly focuses on developing methodologies for large-scale GRN
analyses. It includes applications of a stochastic process theory called G-networks
and a reverse engineering technique for large-scale GRNs. Additionally a series
of bioinformatics techniques was applied in brain tumor data to detect disease
candidate genes along with their large-scale GRNs.
The proposed techniques such as stochastic modeling (bottom-up) and reverse
engineering (top-down) could provide a systematic view of a complex system and
an efficient guideline to identify candidate genes or pathways triggering a specific
phenotype of a cell. As further work, the combinatorial use of the modeling and
reverse engineering approaches would be helpful in obtaining a reliable mathematical model and even in developing a synthetic biological system.