ABSTRACT: BACKGROUND: Inferring the structure of gene regulatory networks (GRN) from a collection of gene expression datahas many potential applications, from the elucidation of complex biological processes to theidentification of potential drug targets. It is however a notoriously difficult problem, for which themany existing methods reach limited accuracy. RESULTS: In this paper, we formulate GRN inference as a sparse regression problem and investigate theperformance of a popular feature selection method, least angle regression (LARS) combined withstability selection, for that purpose. We introduce a novel, robust and accurate scoring technique forstability selection, which improves the performance of feature selection with LARS. The resultingmethod, which we call TIGRESS (for Trustful Inference of Gene REgulation with StabilitySelection), was ranked among the top GRN inference methods in the DREAM5 gene networkinference challenge. In particular, TIGRESS was evaluated to be the best linear regression-basedmethod in the challenge. We investigate in depth the influence of the various parameters of themethod, and show that a fine parameter tuning can lead to significant improvements andstate-of-the-art performance for GRN inference, in both directed and undirected settings. CONCLUSIONS: TIGRESS reaches state-of-the-art performance on benchmark data, including both in silico and invivo (E. coli and S. cerevisiae) networks. This study confirms the potential of feature selectiontechniques for GRN inference. Code and data are available on http://cbio.ensmp.fr/tigress.Moreover, TIGRESS can be run online through the GenePattern platform (GP-DREAM,http://dream.broadinstitute.org).