BENIN: combining knockout data with time series gene expression data for the gene regulatory network inference

Abstract

Gene regulatory network inference is one of the central problems in computational biology. The limited availability of biological data as well as the intrinsic noise they contain have triggered the need of models that integrate the vast variety of data available to take advantage of the complementarity of the information they provide about regulation. With this idea in mind, we propose BENIN: Biologically Enhanced Network INference. BENIN is a general framework that jointly considers prior knowledge with expression data to boost the network inference. This method considers network inference as a feature selection problem. To solve it, BENIN uses a penalized regression method, elastic net, combined with bootstrap resampling. Using the benchmark dataset from the DREAM 4 challenge, we demonstrate that, when using times series expression data with knockout gene expression data, BENIN significantly outperforms other methods.