Inference of Gene Regulatory Network using Factorization of Gene Expression Profile Matrix

초록

The gene regulatory network represents a volatile structure that mediates the flow of information controlling and regulating life. This network is believed to play a key role in life, to lead the evolution of life and the difference between individuals or species, and to explain the pathophysiology of diseases. If is assumed that the expression of genes can be controlled and regulated at the proper time by some logic and connection strength between genes or modules of genes. Because high-throughput data is still insufficient to understand the gene regulatory network, it is necessary to develop a model reflecting the real gene regulatory network and to approach each part of the gene regulatory network based on the model. In this study, a gene regulatory network of differentiating mouse embryonic stem (ES) cells was inferred by obtaining both clusters of genes and channels transmitting regulatory information between the clusters as nodes and edges of the network, respectively. A time-series gene expression dataset were obtained from cDNA microarray experiments at six point of time after induction of neuronal differentiation of mouse ES cells. The eigengenes of each cluster were factorized into two matrices, the gene expression profile matrix and the weight coefficient matrix. By using the significance testing of the weight coefficients, 34 genes that mainly comprise the first eigengens of the clusters were identified. Such genes, defined as the principal genes, were considered to be crucial in regulating the clusters. By putting the results together, a gene regulatory network was inferred, which transmits regulatory information between clusters mediated by the principal genes. The analysis of biological significance of the principal genes and the cluster network supports the feasibility of the algorithms. The clusters that include the principal genes are understood to drive the manifestation of the phenotypes, e.g., the neuronal differentiation of mouse ES cells.