Analysis of four long non-coding RNAs for hepatocellular carcinoma screening and prognosis by the aid of machine learning techniques

Abstract

Hepatocellular carcinoma (HCC) represents a significant health burden in Egypt, largely attributable to the endemic prevalence of hepatitis B and C viruses. Early identification of HCC remains a challenge due to the lack of widespread screening among at-risk populations. The objective of this study was to assess the utility of machine learning in predicting HCC by analyzing the combined expression of lncRNAs and conventional laboratory biomarkers. Plasma levels of four lncRNAs (LINC00152, LINC00853, UCA1, and GAS5) were quantified in a cohort of 52 HCC patients and 30 age-matched controls. The individual diagnostic performance of each lncRNA was assessed using ROC curve analysis. Subsequently, a machine learning model was constructed using Python’s Scikit-learn platform to integrate these lncRNAs with additional clinical laboratory parameters for HCC diagnosis. Individual lncRNAs exhibited moderate diagnostic accuracy, with sensitivity and specificity ranging from 60 to 83% and 53–67%, respectively. In contrast, the machine learning model demonstrated superior performance, achieving 100% sensitivity and 97% specificity. Notably, a higher LINC00152 to GAS5 expression ratio significantly correlated with increased mortality risk. The integration of lncRNA biomarkers with conventional laboratory data within a machine learning framework demonstrates significant potential for developing a precise and cost-effective diagnostic tool for HCC. To enhance the model’s robustness and prognostic capabilities, future studies should incorporate larger cohorts and explore a wider array of lncRNAs.