Two Statistical Models for Biomolecular Network Evolution

Abstract

Two Statistical Models for Biomolecular Network Evolution Chen-Hsiang Yeang Institute of Statistical Science, Academia Sinica, Taiwan ?: There are a plethora of statistical models to study the sequence evolution of single genes or loci.? However, relatively few such models exist in the arena of network evolution.? In this talk I will introduce two statistical models facilitating the characterization and comprehension of large-scale biomolecular network evolution.? In the first part, I will describe a probabilistic graphical model that reconstructs the evolutionary history of domain architectures and catalytic functions of enzyme proteins.? We applied this model to the metabolic systems of 13 prokaryotic and eukaryotic species and discovered general characteristics and guiding principles of metabolic network evolution.? In the second part, I will describe a birth-death process model for the evolution of cis-regulatory elements in gene regulatory networks.? An algorithm to estimate the strength of natural selection of these elements according to the birth-death model was proposed.? We incurred a large-scale screening of all the 10-nucleotide sequences on 34 mammalian species, and found that selection coefficients of candidate cis-regulatory elements were strongly associated with their functional relevance.? In both parts I will cover multiple aspects of bioinformatic modeling including mathematical formulation, computation, validation and biological insights.