Explainable Biomarker Discovery in Hepatocellular Carcinoma: an XGBoost and SHAP Framework

Hepatocellular carcinoma (HCC) is one of the leading causes of cancer-related deaths worldwide. Although ensemble machine learning models have demonstrated high accuracy in HCC classification and prediction, their "black-box" nature impedes their reliable adoption in clinical practice. Therefore, this study proposes a two-stage research framework: first, constructing a high-accuracy HCC risk prediction model using the extreme gradient boosting algorithm; second, applying the SHAP method to provide global and local interpretations of the model's predictions, quantifying the contribution of each feature to the predictive outcome, thereby identifying key features. The research is based on the HCC clinical feature dataset, while also drawing on experiences from multi-center databases such as TCGA. This work aims to provide clinicians with transparent decision support, promote personalized treatment, and offer clear target maps for subsequent biological validation studies.