Mining gene regulatory networks by neural modeling of expression timeseries

Abstract

Several machine learning techniques have been developed for discovering interesting and unknown relations between variables from data, even more when these techniques can assist in understanding the behaviour of a complex system. This behaviour can be represented by the interactions between its variables, for instance as a directed graph. A gene regulatory network (GRN) is an abstract mapping of gene regulations in living organisms that can help to predict the system behavior. During last years, many approaches have been proposed to unravel the complexity of gene regulation. Genes interact with one another and these interactions can be measured over a number of time steps, producing temporal gene expression profiles. A hot topic on gene expression data analysis nowadays is the reconstruction of a GRN from such data, revealing the underlying network of genetogene interactions. In other words, the goal is to determine the pattern of activations and inhibitions among genes that make up the underlying GRN.