PLAC8 Variant‐Informed Multi‐Omics and AI Model for Survival Prediction in Advanced HCC Treated With Radiotherapy Plus Targeted Therapy and PD‐1 Inhibitors
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
Background Radiotherapy combined with immune‐targeted therapy is increasingly used for advanced HCC, but outcomes remain heterogeneous due to substantial clinical and molecular diversity, and robust tools for individualized risk stratification are lacking. Methods We retrospectively analyzed 150 patients who received intensity‐modulated radiotherapy plus PD‐1 inhibitors and targeted therapy. Pretreatment contrast‐enhanced CT images were used to delineate the gross tumor volume as the region of interest. A total of 1130 radiomics features were extracted. LASSO regression was applied to derive a radiomics‐based risk score and portal vein tumor thrombosis (PVTT), tumor size, and serum alpha‐fetoprotein (AFP) level. In parallel, multi‐omics analyses were performed to identify core genes and biological pathways, and PLAC8 was determined as the final hub gene for downstream biological interpretation and validation. Model performance was assessed by C‐index, time‐dependent receiver operating characteristic (ROC) curves, calibration, decision curve analysis, and SHAP‐based interpretation. Results Median overall survival (mOS) in the entire cohort was 29.2 months. In the training set, C‐indices for the Cox, RSF, and XGBoost models were 0.756, 0.753, and 0.831, respectively, indicating superior discrimination for the XGBoost model. In the validation cohort, the XGBoost model yielded 1‐, 2‐, and 3‐year time‐dependent areas under the ROC for OS of 0.795, 0.848, and 0.769, respectively, with good calibration and greater net clinical benefit than either treat‐all or treat‐none strategies. SHAP analysis identified AFP as the dominant contributor to model predictions, followed by tumor size, the radiomics risk score, and PVTT. In parallel, multi‐omics analyses revealed enrichment of cell cycle‐related programs and prioritized PLAC8 as the final hub gene. Genetic analyses further supported a PLAC8‐locus variant–expression–immune infiltration axis, providing complementary biological context for the imaging‐ and clinic‐based risk stratification framework. Conclusion The LASSO–XGBoost model enables clinically applicable individualized survival prediction for advanced HCC treated with radiotherapy plus targeted therapy and PD‐1 inhibitors, while multi‐omics findings—highlighted by PLAC8—provide additional biological insight into outcome heterogeneity.
Fields of Study & Subjects
Hepatocellular Carcinoma Treatment and Prognosis
Radiomics and Machine Learning in Medical Imaging
Cancer Immunotherapy and Biomarkers
Publication details
- Journal / source
- Human Mutation
- Volume
- 2026
- Issue
- 1
- Pages
- 8878584
- Publication date
- 1 January 2026
- Language
- EN