Gastric cancer (GC) is the third leading cause of cancer deaths and the fourth most prevalent malignancy worldwide. The high incidence and mortality rates of gastric cancer result from multiple factors such as ineffective screening, diagnosis, and limited treatment options. In our study, we sought to systematically identify predictive molecular networks and key regulators to elucidate complex interacting signaling pathways in GC. We performed an integrative network analysis of the transcriptomic data in The Cancer Genome Atlas (TCGA) gastric cancer cohort and then comprehensively characterized the predictive subnetworks and key regulators by the matched genetic and epigenetic data. We identified 221 gene subnetworks (modules) in GC. The most prognostic subnetworks captured multiple aspects of the tumor microenvironment in GC involving interactions among stromal, epithelial and immune cells. We revealed the genetic and epigenetic underpinnings of those subnetworks and their key transcriptional regulators. We computationally predicted and experimentally validated specific mechanisms of anticancer effects of GKN2 in gastric cancer proliferation and invasion in vitro. The network models and the key regulators of the tumor microenvironment in GC identified here pave a way for developing novel therapeutic strategies for GC.