报告题目（Title）：随机网络的概率景观和概率流的精确构造 "Exact construction of probability landscapes and global flow maps of discrete flux of stochastic networks"
报 告 人（Speaker）：梁杰教授 (伊利诺大学芝加哥分校生物工程系 )
报告时间（Time）： 2019年10月29日（周二）13:00-14:30
报告地点（Place）：校本部G309
邀请人（Inviter）：敖平
摘要（Abstract）： The discrete Chemical Master Equation (dCME) provides a fundamental framework to study stochasticity in reaction networks. However, solving dCMEs is challenging, as it is often unknown if the computed probability peaks are complete and nothing important is missing. It is also difficult to accurately estimate probabilities of rare events. Here we discuss how time-evolving and steady state probability landscape can be exactly computed using the n-simplex optimal state enumeration algorithm and the ACME (accurate chemical master equation) method (www.bioacme.org, DOI. 10.1137/15M1034180), without Monte Carlo simulation or Fokker-Planck/Langevin approximation. We also discuss how to guarantee small errors and how to estimate the best achievable accuracy with a given laptop/supercomputer, through an a priori calculated error bound when truncation of the state space inevitabley occurs for complex networks. In addition, we describe a new formulation of discrete probability flux and velocity under proper boundary conditions and how their global flow maps can be exactly computed. We give examples on analysis of phenomenological characterization of cellular decision networks, including bi-stability, epigenetic states, and the robustness of wild type versus mutants networks. Furthermore, we discuss how ACME computation can reveal the origin of the reservoir of HIV latently infected cells and uncover detailed mechanism of probabilistic intra-cellular control of latency and transactivation. Moreover, we discuss how new approaches for health intervention can been formulated based on large scale computation exploring changes in the probability landscapes of networks under different perturbations, in the context of clinical treatment of the "shock and kill" and the "block and lock" strategies.
