Abstract
BACKGROUND: Altered glycosylation, one of the most common post-translational protein modifications, plays a critical role in the initiation and progression of soft tissue sarcoma (STS). Dysregulated expression of glycosyltransferases leads to abnormal glycosylation patterns, which may offer valuable insights for prognosis and therapeutic response prediction in STS. METHODS: Transcriptional variants and expression profiles of glycosylation-related genes were analyzed using data from The Cancer Genome Atlas (TCGA). Differential gene expression analysis and non-negative matrix factorization (NMF) were performed to identify STS molecular subtypes. A comprehensive machine learning framework integrating 101 algorithms was applied to construct a glycosyltransferase-based prognostic signature. Kaplan-Meier analysis, Cox regression, and receiver operating characteristic (ROC) curves were used to assess the prognostic value of the model. Immune infiltration was evaluated using multiple computational approaches, and functional validation was conducted via in vitro experiments. RESULTS: Two distinct STS subtypes with significant immunological and clinical differences were identified. A 12-gene glycosyltransferase signature was developed, effectively stratifying patients into high-risk and low-risk groups based on the median riskscore. The high-risk group demonstrated significantly poorer survival outcomes. Immune profiling revealed greater immunosuppression in the high-risk group. In vitro silencing of STT3A significantly suppressed proliferation and migration of STS cells. CONCLUSIONS: The proposed glycosylation-related gene signature accurately distinguishes between high- and low-risk STS patients and may serve as a reliable prognostic tool. It also provides novel insights into tumor immune microenvironment and potential therapeutic targets for STS.