IntroductionA deadly gynecological cancer, ovarian cancer (OV), has a poor prognosis because of late-stage diagnosis and few targeted therapies. Addressing the tumor microenvironment (TME) in solid tumors has shown promise since it is crucial in promoting cancer progression. MethodsWe obtained bulk RNA-seq data from TCGA-OV, GSE26712, GSE102073, and ICGC cohorts, as well as scRNA-seq data from EMTAB8107, GSE118828, GSE130000, and GSE154600 cohorts using the TISCH2 database. The ConsensusClusterPlus package was used to cluster the OV tumor tissues hierarchically to determine two molecularly different groups (C1 and C2). A total of ten different types of machine learning techniques with 101 combinations were used for prognostic model construction. Using eight TME algorithms integrated into the IOBR R package, the bulk RNA-seq dataset was analyzed. For in vitro experiments, OVCAR3 and SKOV3, two OV cell lines, were used. The migratory potential of the ovarian cancer cells was assessed using Transwell assay, while proliferation was assessed using CCK8 assay. ResultsBased on TME-related gene set expression, two distinct molecular subgroups (C1 and C2) were identified through consensus clustering, with C1 showing higher TME activity. Further analysis indicated that C1 had increased cancer-associated fibroblasts (CAFs), M1 macrophages, and CD8+ T cells, suggesting a more activated and pro-inflammatory TME. Drug sensitivity analysis revealed that 5-Fluorouracil might be beneficial to C1 patients. Functional differences between C1 and C2 were identified, including cell adhesion, mononuclear cell differentiation, and leukocyte migration. A machine learning model was developed to create a TME-related prognostic signature, demonstrating strong prognostic capabilities across multiple datasets. High-risk patients showed a more immune-suppressive TME and higher tumor stemness. ScRNA-seq disclosed a highly activated TME-related signature in OV. Cancer cell lines had significantly higher SH2D1A mRNA expression than normal ovarian epithelial cells. We observed that SH2D1A knockdown in 2 ovarian cancer cell lines (OVCAR3 and SKOV3) reduced migration and proliferation through a series of in-vitro experiments. ConclusionTME-associated genes were efficient in ovarian cancer molecular subtyping. A TME-based prognosis model was constructed for vigorous prognostic stratification efficacy across multiple datasets. Moreover, we identified a pivotal role of SH2D1A in promoting proliferation and migration in ovarian cancer.