A new method for identifying cancer-related gene association patterns

Abstract

Gene association plays important roles in complex genetic pathology of cancer. However, development of methods for finding cancer-related gene associations is still in its infancy. Based on a biological concept of gene association module (GAM) comprising a center gene and its expression-related genes, this paper proposes a gene association detection model called kernel GAM (kGAM). In the model, we assume that the expression of the center gene can be predicted by the expression-related genes. Based on defining a cost function, a kernel ridge regression algorithm is developed to solve the kGAM model. Finally, to identify a compact GAM for a given center gene, a heuristic search procedure is designed. Experimental results on three publicly available gene expression data sets show the effectiveness and efficiency of the proposed kGAM model in identifying cancer-related gene association patterns. © 2012 Springer-Verlag.