A Hybrid XGBoost-MLP Approach for Predicting miRNA-Disease Associations

The discovery of miRNA-disease connections is essential for furthering biomedical research since microRNAs (miRNAs) control gene expression and are linked to a wide range of disorders. This paper proposes the use of a hybrid machine learning strategy that couples the strengths of Multi-Layer Perceptron, for classification, with XGBoost, for feature extraction, toward the prediction of miRNA-disease connections. Complex interactions between features are captured in an efficient way by XGBoost and then classified by the MLP on the final data after processing the gathered features. When compared to standalone models, the model performed better when measured by accuracy, precision, recall, and F1 score. Furthermore, in order to improve the model's interpretability, we utilized SHAP (SHapley Additive exPlanations), which quantifies the effect of each miRNA on illness predictions in order to shed light on feature contributions.