Machine learning-based integration develops a neutrophil-derived signature for improving outcomes in hepatocellular carcinoma.

Abstract

The heterogeneity of tumor immune microenvironments is a major factor in poor prognosis among hepatocellular carcinoma (HCC) patients. Neutrophils have been identified as playing a critical role in the immune microenvironment of HCC based on recent single-cell studies. However, there is still a need to stratify HCC patients based on neutrophil heterogeneity. Therefore, developing an approach that efficiently describes "neutrophil characteristics" in HCC patients is crucial to guide clinical decision-making. We stratified two cohorts of HCC patients into molecular subtypes associated with neutrophils using bulk-sequencing and single-cell sequencing data. Additionally, we constructed a new risk model by integrating machine learning analysis from 101 prediction models. We compared the biological and molecular features among patient subgroups to assess the model's effectiveness. Furthermore, an essential gene identified in this study was validated through molecular biology experiments. We stratified patients with HCC into subtypes that exhibited significant differences in prognosis, clinical pathological characteristics, inflammation-related pathways, levels of immune infiltration, and expression levels of immune genes. Furthermore, A risk model called the "neutrophil-derived signature" (NDS) was constructed using machine learning, consisting of 10 essential genes. The NDS's RiskScore demonstrated superior accuracy to clinical variables and correlated with higher malignancy degrees. RiskScore was an independent prognostic factor for overall survival and showed predictive value for HCC patient prognosis. Additionally, we observed associations between RiskScore and the efficacy of immune therapy and chemotherapy drugs. Our study highlights the critical role of neutrophils in the tumor microenvironment of HCC. The developed NDS is a powerful tool for assessing the risk and clinical treatment of HCC. Furthermore, we identified and analyzed the feasibility of the critical gene RTN3 in NDS as a molecular marker for HCC.