Modeling Gene Expression Dynamics by Kernel Auto-RegressiveModels for Time-Course Microarray Data

The DNA microarray technology has shown its extensive applications in clinical research and has emerged as a powerful tool for understanding gene expressions through a simultaneous study of thousands of genes. A successful modeling of gene profiles can provide a pathway of revealing gene regulations from the microarray data. Therefore, modeling the gene expression networks has attracted increasing interests in computational biology community. We propose a nonlinear dynamical system based on kernel auto-regressive model in this application. The proposed method can analyze the nonlinear mapping among the gene expression dynamics by using the kernels. A sparse model is employed so as to decrease the computational cost and improve the illustration ability of the method. We use the kernel recursive least squares, which is an approximation of the kernel principal component analysis, in building the sparse model. By presenting simulation results, we show that dynamical nonlinear networks are attractive and suitable for modeling gene expression profiles. A range of challenging research problems will also be discussed in this paper.