The availability and utility of genome-scale metabolic networks have exploded with modern genome-sequencing capabilities. However, these generic models overlooked actual physiological states of the tissues and included all the reactions implied by the genome annotations. To address this problem, we reconstructed a human heart-specific metabolic network based on transcriptome and proteome data. The resulting model consists of 2803 reactions and 1880 metabolites, which correspond to 1721 active enzymes in human heart. Using the model, we detected 24 epistatic interactions in human heart, which are useful in understanding both the structure and function of cardiovascular systems. In addition, a set of 776 potential biomarkers for cardiovascular disease (CVD) has been successfully explored, whose concentration is predicted to be either elevated or reduced because of 278 possible dysfunctional cardiovascular-associated genes. The model could also be applied in predicting selective drug targets for eight subtypes of CVD. The human heart-specific model provides valuable information for the studies of cardiac activity and development of CVD. (C) 2011 Elsevier Inc. All rights reserved.