A machine learning-based immune response signature to facilitate prognosis prediction in patients with endometrial cancer

Endometrial cancer is the most prevalent form of gynecologic malignancy, with a significant surge in incidence among youngsters. Although the advent of the immunotherapy era has profoundly improved patient outcomes, not all patients benefit from immunotherapy; some patients experience hyperprogression while on immunotherapy. Hence, there is a pressing need to further delineate the distinct immune response profiles in patients with endometrial cancer to enhance prognosis prediction and facilitate the prediction of immunotherapeutic responses. The ssGSEA method was used to evaluate the activities of the immune response pathways in patients with endometrial cancer. Unsupervised clustering was employed to identify the different immune response patterns. WGCNA was employed to identify the genes that highly correlated with the immune response patterns observed. Ninety-five machine learning combinations were utilized to identify the optimal prognosis model and the novel biomarker, SLC38A3. Experiments such as cell invasion, migration, scratch, and in vivo tumorigenicity were performed to determine the function of SLC28A3. Molecular docking techniques were employed to determine the targeted action of periodate-oxidized adenosine on SLC38A3. Patients exhibited both immune response-suppressing C1 phenotypes and immune response-activating C2 phenotypes, with significant differences in prognosis between these two phenotypes. WGCNA identified 418 genes that highly correlated with the immune response phenotypes, of which 69 genes were associated with prognosis. The immune response-related score (IRRS) established by multiple machine learning frameworks demonstrated stability in predicting patient prognosis and immune status. High expression of SLC38A3 contributes to cellular malignant traits, and periodate-oxidized adenosine bound stably to SLC38A3. IRRS accurately predicts disease prognosis and immune status in patients with endometrial cancer. SLC38A3 serves as a prognostic marker for these patients and can be stably targeted by periodate-oxidized adenosine.