Simple Summary Soft tissue sarcomas (STS) still lack effective clinical stratification and prognostic models. The aim of this study is to establish a reliable prognostic gene signature in STS. Using 189 STS samples from the TCGA database, a four-gene signature (including DHRS3, JRK, TARDBP and TTC3) and nomograms that can be used to predict the overall survival and relapse free survival of STS patients was developed. The predictive ability for metastasis free survival was externally verified in the GEO cohort. We demonstrated that the novel gene signature provides an attractive platform for risk stratification and prognosis prediction of STS patients, which is of great importance for individualized clinical treatment and long-term management of patients with this rare and severe disease. Abstract Soft tissue sarcomas (STS), a group of rare malignant tumours with high tissue heterogeneity, still lack effective clinical stratification and prognostic models. Therefore, we conducted this study to establish a reliable prognostic gene signature. Using 189 STS patients’ data from The Cancer Genome Atlas database, a four-gene signature including DHRS3, JRK, TARDBP and TTC3 was established. A risk score based on this gene signature was able to divide STS patients into a low-risk and a high-risk group. The latter had significantly worse overall survival (OS) and relapse free survival (RFS), and Cox regression analyses showed that the risk score is an independent prognostic factor. Nomograms containing the four-gene signature have also been established and have been verified through calibration curves. In addition, the predictive ability of this four-gene signature for STS metastasis free survival was verified in an independent cohort (309 STS patients from the Gene Expression Omnibus database). Finally, Gene Set Enrichment Analysis indicated that the four-gene signature may be related to some pathways associated with tumorigenesis, growth, and metastasis. In conclusion, our study establishes a novel four-gene signature and clinically feasible nomograms to predict the OS and RFS. This can help personalized treatment decisions, long-term patient management, and possible future development of targeted therapy.