Potential outcomes from simulation and analysis of genome-scale models (5). Like any modeling framework, constraint-based models can lead to either accurate or inaccurate predictions. In cases in which an accurate result is found, microarray data have been used to validate the model and methods of analysis, provide data for potential insight into the biology reflected by the simulations, or suggest simplifications for an otherwise complex biological system. For example, as depicted in the graph, E. coli metabolism and its associated TRN exhibit few functional modes, affected mostly by the electron acceptor used. Activity profiles formed three clusters, separated as a function of the electron acceptor (represented by the ellipses), the carbon source (represented by separation inside each ellipse), and to a lesser degree, the nitrogen source. If, however, the simulations yield inaccurate predictions, novel findings can be obtained through alternative analytical methods. Predictions of inactive pathways, novel pathways, or new regulatory rules are just examples of new knowledge that may be gained this way. DMSO, dimethyl sulfoxide; TMAO, trimethylamine N- oxide.

Pipeline of data to generate genome-scale in silico models that are predictive. Constraint-based modeling is evolving into a process with two steps in computation: (i) reconstruction and (ii) simulation. Novel computational methods are able to integrate genome sequences and various high-throughput data types (gene expression data and chromatin immunoprecipitation-on-chip/chromatin immunoprecipitation-sequencing data, etc.) to aid in the reconstruction of metabolic and transcription/translation networks and TRNs. After curation, an integrated network reconstruction can be converted into a model based on the organism's genotype and environmental conditions. From that point, hypotheses can be tested in silico and exploratory analysis can be conducted. WT, wild-type.