Research Project
10268260
PROJECT SUMMARY/ABSTRACT Cellular state transitions (i.e. from pluripotent to committed, replicative to quiescent, etc.) require the coordinated regulation of thousands of genes. Therapeutically harnessing these transitions holds great promise for human health;? for instance, autologous stem cell therapy has been successfully used in regenerative medicine and cancer treatments, among others. While some of the key regulatory switches are known, the field lacks a systems-level understanding of the genomic circuits that control these transitions, information that is critical for informed clinical intervention. Here, we will develop an integrated computational framework to identify core gene regulatory circuits from large gene networks and predict their dynamics and regulatory functions without the need of detailed network kinetic parameters. Advances in genomics profiling technology have enabled the mapping of gene regulatory networks, thus we now have the capacity to identify combinatorial interactions among genes and the master regulators of state transitions. Some systems biology approaches have simulated the dynamics of a gene regulatory circuit, but traditional methods suffer from two key issues. First, there is no rational rule to choose an appropriate set of regulator genes in a large network to model. Second, since it is hard to directly measure most network kinetic parameters from experiment, modeling results are based on a set of guessed parameters that can be less than optimal, limiting the application of mathematical modeling to large systems and the prediction power of systems biology. To address these issues, we recently developed algorithm named random circuit perturbation (RACIPE). RACIPE generates an ensemble of circuit models, each of which corresponds to a distinct set of random kinetic parameters, and uniquely identifies robust features, such as clusters of stable gene expression states, by statistical analysis. We will further enhance RACIPE algorithms for large systems and new data analysis tools using machine learning. This approach will convert a traditional nonlinear dynamics problem into a data analysis problem, an essential step for extending the application of gene circuit modeling to large systems. It also provides a novel strategy to integrate top-down genomics approaches with bottom-up mathematical modeling. The algorithms will be tested and refined using literature-based gene networks, public genomics data, and data from collaboration, with a focus on cell differentiation in developmental processes and state transitions in oncogenesis. Success of the project will result in a comprehensive toolkit that will unveil the gene regulatory mechanism of cellular decision-making in any cell of interest. The algorithmic development is expected to have a broad impact on not only basic research in systems biology but also shed light on therapeutic intervention in genomic medicine.
