Abstract
Tumor stemness has been reported to play important roles in cancers. However, a comprehensive analysis of tumor stemness remains to be performed to investigate the specific mechanisms and practical values of stemness in soft tissue sarcomas (STS). Here, we applied machine learning to muti-omic data of patients from TCGA-SARC and GSE21050 cohorts to reveal important roles of stemness in STS. We demonstrated limited roles of existing mRNAsi in clinical application. Therefore, based on stemness-related signatures (SRSs), we identified three stemness subtypes with distinct stemness, immune, and metabolic characteristics using consensus clustering. The low-stemness subtype had better prognosis, activated innate and adaptive immunity (e.g., infiltrating B, DC, Th1, CD8+ T, activated NK, gamma delta T cells, and M1 macrophages), more enrichment of metabolic pathways, more sites with higher methylation level, higher gene mutations, CNA burdens, and immunogenicity indicators. Furthermore, the 16 SRS-based stemness prognostic index (SPi) was developed, and we found that low-SPi patients with low stemness had better prognosis and other characteristics similar to those in the low-stemness subtype. Besides, low-stemness subtype and low-SPi patients could benefit from immunotherapy. The predictive value of SPi in immunotherapy was more accurate after the addition of MSI into SPi. MSI(low)SPi(low) patients might be more sensitive to immunotherapy. In conclusion, we highlighted mechanisms and practical values of the stemness in STS. We also recommended the combination of MSI and SPi which is a promising tool to predict prognosis and achieve precise treatments of immunotherapy in STS.