Network Discovery in Equilibrium-state and Dynamic Data:Applications to Phosphoproteomics and Kinetics.

Network Discovery;Phosphoproteomics;Kinetics;Equilibrium-state Data;Dynamic Data;Filamentous Growth;Science (General);Science;Bioinformatics

Zhang, YanSchnell, Santiago David ;

University of Michigan

关键词: Network Discovery;    Phosphoproteomics;    Kinetics;    Equilibrium-state Data;    Dynamic Data;    Filamentous Growth;    Science (General);    Science;    Bioinformatics;   

Others  :  https://deepblue.lib.umich.edu/bitstream/handle/2027.42/96061/zhyan_1.pdf?sequence=1&isAllowed=y

瑞士|英语

来源: The Illinois Digital Environment for Access to Learning and Scholarship

【 摘 要 】

My research has focused on network discovery from phosphoproteomics and kinetics data. My work contains methodology development and evaluation, and their applications to large-scale datasets. Specifically, Chapter 2 proposes a quantitative analysis pipeline for equilibrium-state phosphoproteomics data; Chapter 3 presents a general method for analyzing time-course kinetics data, using a differential equation-based method; and Chapter 4 presents an approach combining Bayesian network structure learning with time-delay detection to analyze time-course data. In Chapter 2, I present a new comprehensive quantitative analysis pipeline for systematic network discovery from equilibrium-state interventional phosphorylation data. The purpose is identification of key proteins in specific pathways, discovering protein-protein relationships, and inferring signaling networks. I also made an effort to partially compensate for the missing value issue. The pipeline is successfully applied to interventional experiments identifying phosphorylation events underlying the transition to filamentous growth in yeast. Five of our predicted proteins were tested by phenotypic experiments, and they all present differential invasive growth, providing validation for our approach.In Chapter 3, I present a computational method, MIKANA Ver. 2, which integrates mathematical modeling of biochemical networks with statistical methods to infer kinetic network structure and estimate reaction parameters from time-course data. It contains multiple improvements and extensions to MIKANA Ver. 1. MIKANA Ver. 2 newly supports autocatalytic reactions and third-order reactions. Penalized regression coupled with non-linear parameter fitting has been used to optimize both oscillatory and linear systems. The prediction precision and stability has been improved on simple models, including second-order oscillatory models. Although MIKANA Ver. 2 improves deficiencies of the previous version, it still has limitations which are discussed in my thesis.In Chapter 4, I present another method for network discovery from time-course data.The novelty of the method presented in this chapter is the combination of Bayesian structure learning with time-delay detection. The time-delay patterns might be easily missed by correlation coefficient-based cluster analysis. I also sought to identify the critical number of time points, at and above which the method can obtain higher-confidence network structures.The new method is applicable to various dynamics data including phosphorylation dynamics in response to specific stimuli.