Identification of metabolic network models from incomplete high-throughput datasets

• DOI: 10.1093/bioinformatics/btr225
• Publication Date: 2011-06-17T23:32:32Z
• Authors: Sara Berthoumieux; Matteo Brilli; Hidde de Jong; Daniel Kahn; Eugenio Cinquemani
• Keywords: Identification; Maximization

Abstract Motivation: High-throughput measurement techniques for metabolism and gene expression provide a wealth of information the identification metabolic network models. Yet, missing observations scattered over dataset restrict number effectively available datapoints make classical regression inaccurate or inapplicable. Thorough exploitation data by that explicitly cope with is therefore major importance. Results: We develop maximum-likelihood approach estimation unknown parameters models relies on integration statistical priors to compensate data. In context linlog modeling framework, we implement method an Expectation-Maximization (EM) algorithm simpler direct numerical optimization method. evaluate performance our methods comparison existing approaches, show EM provides best results variety simulated scenarios. then apply real problem, model Escherichia coli central carbon metabolism, based challenging experimental from literature. This leads promising allows us highlight critical issues. Contact: [e-mail]; [e-mail] Supplementary information: are at Bioinformatics online.