Needle in a haystack: effective search for dynamics across the space of networks

点击翻译按钮获取中文摘要

This proposal introduces a revolutionary new approach that allows provably accurate description of network dynamics that is valid for all initial data and parameters. The result of the computation is encoded as a searchable database of Dynamic Signatures of Gene Regulatory Networks (DSGRN). The proposal develops mathematical theory and computational tools to efficiently compute and interrogate the database for biologically important correlates of network function. This approach will be used to search space of networks for experimentally observed dynamics related to cell cycle dynamics in two types of yeast (S. cerevisiae and Cryptococcus neoformans), and humans. In Aim 1, DSG RN is used to find perturbations in parameter space that lead from a cancer to a benign phenotype in a human cell cycle restriction point network. Close homology of the human network with an analogous network in S. cerevisiae will be used to experimentally verify our findings. Aim 2 takes advantage of DSGRN ability to search space of networks to (a) find cell cycle network for C. neoformans, a causal agent of a deadly fungal infection, and (b) find potential missing links in human cell cycle networks. Detailed understanding of human genome and cellular networks has not yet led to a quantum leap in medicine due to complex relationship of the genome with cell phenotypes. This gap is manifested in mathematical context by the lack of methods that predict dynamics of a network, at all parameters, from its structure. This proposal addresses this gap by developing modeling and computational approach that allows provably accurate description of dynamics that is valid for all initial data and parameters. The description of the dynamics is coarser than the description by standard models. While DSG RN does not have the acuity at the level of single trajectories, the database can be queried for bistability, recurrent nonequilibrium dynamics, particular steady state expression patterns, and patterns of extrema that match experimental time series data. The coarse characterization of dynamics allows searching thousands and potentially millions of networks for the desired coarse dynamics, and, if needed, to refine the model to higher acuity for the networks and parameter regions that exhibit this dynamics.