Modularization of physiological properties of gene expression in hES cell differentiation

초록

Under multiple causative environmental and internal conditions, a cell can build a scenario-based system that is the combination of gene modules with physiological functions. The interconnection of gene modules within the system contains information flow and determines the cellular behaviors by the effects of the quantity and route of information. The gene modularization is the first step for outlining the operations of a vast genome space. Also, high throughput gene expression profiling is one of the most important approaches to enhance the understanding of complex physiological functions. A specific cellular feature represented by phenotypic and genotypic characteristics of cells is responsible for the gene expression patterns, which are shared by a group of genes, i.e., a gene module. Therefore, searching gene modules from high-throughput gene expression data can deduce the significance results for physiological properties of cell. Our proposed algorithm differs from the existing algorithm in that a gene module is not clustered according to the similarity or dissimilarity of expression patterns. In this research, a gene module is a single unit with linear property generated by matrix decomposition method. In addition, we suggest a modified genetic algorithm (GA) based method for exploring gene modules. The modified GA uses as a local search method for the optimization of total system. During repetitive iteration of the modified GA, a randomly generated set of gene modules was converged to an optimal set. This method was applied to cDNA microarray gene expression data obtained from the human embryonic stem (hES) cells differentiation to neuron. As a result, globally optimized gene modules showed consistent physiological properties with our knowledge of stem cell differentiation.