A stemness-based signature with inspiring indications in discriminating the prognosis, immune response, and somatic mutation of endometrial cancer patients revealed by machine learning.

Endometrial cancer (EC) is a fatal gynecologic tumor. Bioinformatic tools are increasingly developed to screen out molecular targets related to EC. Our study aimed to identify stemness-related prognostic biomarkers for new therapeutic strategies in EC. In this study, we explored the prognostic value of cancer stem cells (CSCs), characterized by self-renewal and unlimited proliferation, and its correlation with immune infiltrates in EC. Transcriptome and somatic mutation profiles of EC were downloaded from TCGA database. Based on their stemness signature and DEGs, EC patients were divided into two subtypes via consensus clustering, and patients in Stemness Subtype I presented significantly better OS and DFS than Stemness Subtype II. Subtype I also displayed better clinicopathological features, and genomic variations demonstrated different somatic mutation from subtype II. Additionally, two stemness subtypes had distinct tumor immune microenvironment patterns. In the end, three machine learning algorithms were applied to construct a 7-gene stemness subtype risk model, which were further validated in an external independent EC cohort in our hospital. This novel stemness-based classification could provide a promising prognostic predictor for EC and may guide physicians in selecting potential responders for preferential use of immunotherapy. This novel stemness-dependent classification method has high value in predicting the prognosis, and also provides a reference for clinicians in selecting sensitive immunotherapy methods for EC patients.