A Molecular Hepatocellular Carcinoma Prognostic Score System Precisely Predicts Overall Survival of Hepatocellular Carcinoma Patients

Background and aims: With high rates of recurrence post-treatment, hepatocellular carcinoma (HCC) is one of the most common types of cancer worldwide and the major cause of cancer death. To improve the overall survival of HCC patients, identification of a reliable biomarker and precise early diagnosis of HCC remain major unsolved problems.

Methods: We initially screened data from the Cancer Genome Atlas liver cancer cohort to identify potential prognosis-related genes. Then, a meta-analysis of five international HCC cohorts was implemented to validate such genes. Subsequently, artificial intelligence models (random forest and neural network) were trained to predict prognosis accurately, and a log-rank test was performed for validation. Finally, the correlation between the molecular hepatocellular carcinoma prognostic score (mHPS) and the stromal and immune scoring in HCC were explored.

Results: A comprehensive list of 65 prognosis-related genes was obtained, most of which have been not extensively studied thus far. A universal HCC mHPS system depending on the expression pattern of only 23 genes was established. The mHPS system had general applicability to HCC patients (log-rank p<0.05) in a platform-independent manner (RNA sequencing or microarray). The mHPS was also correlated with the stromal and immune scoring in HCC, reflecting the status of the tumor immune microenvironment.

Conclusions: Overall, the mHPS is an easy and cost-effective prognosis predicting system, which can disclose previously uncovered heterogeneity among patient subpopulations. The mHPS system can further stratify patients who are at the same clinical stage and should be valuable for precise treatment. Moreover, the prognosis-related genes recognized in this study have potential in targeted and immune therapy.