Explaining gene responses by linear modeling

Abstract

Increasing our knowledge about molecular processes in response to a certain treatment or infection in plants, insects, or other organisms requires the identification of the genes involved in this response. In this paper, we propose the Profile Interaction Finder (PIF) to identify such genes from gene expression data which is based on a convex linear model, and we investigate its efficacy for two applications related to stimulus response. First, we seek to identify sets of putative regulatory genes that explain the expression levels of a gene under different stimuli best. Second, we aim at identifying genes that show a specific response to a stimulus or a combination of stimuli. For both applications, we study the expression response of two Arabidopsis species to treatment with the plant hormone auxin and of Apis mellifera to pathogen infection. The proposed approach may be of general utility for analyzing expression data with a focus on genes and gene sets that explain specific stimulus response.