Prediction of patients’ survivability in Uterine Cancer using machine learning models

Uterine cancer is a serious worry for women all over the world, and we have used multi-omics datasets to present a model that predicts the survival rate of uterine cancer patients by combining machine learning approaches with eXplainable artificial intelligence (XAI). Total 483 patient samples were collected from The Cancer Genome Atlas database. SNP mutation data, expression data of miRNA and mRNA data were used to build the model. The MLP-MCFS model with the mRNA dataset outperformed other models, with a remarkable accuracy of 97.67%, a Matthews Correlation Coefficient of 91.11%, and an Area Under Receiver Operator Characteristic curve of 98%. In addition, XAI-based SHAP results increased the explainability of the models and extracted top biomarkers that play a significant role in predicting the survivability of uterine cancer patients. The findings may help to determine the survivability of uterine cancer patients after the initial diagnosis with the help of specific top biomarkers.