Hepatocellular carcinoma (HCC) ranks among the most prevalent cancers and accounts for a significant proportion of cancer-associated deaths worldwide. This disease, marked by multifaceted etiology, often poses diagnostic challenges. Finding a reliable and non-invasive diagnostic method seems to be necessary. In this study, we analyzed the gene expression profiles of 20 HCC patients, 12 individuals with chronic hepatitis, and 15 healthy controls. Enrichment analysis revealed that platelet aggregation, secretory granule lumen, and G-protein-coupled purinergic nucleotide receptor activity were common biological processes, cellular components, and molecular function in HCC and chronic hepatitis B (CHB) compared to healthy controls, respectively. Furthermore, pathway analysis demonstrated that “estrogen response” was involved in the pathogenesis of HCC and CHB conditions, while, “apoptosis” and “coagulation” pathways were specific for HCC. Employing computational feature selection and logistic regression classification, we identified candidate genes pivotal for diagnostic panel development and evaluated the performance of these panels. Subsequent machine learning evaluations assessed these panels’ performance in an independent cohort. Remarkably, a 3-marker panel, comprising RANSE2, TNF-α, and MAP3K7, demonstrated the best performance in qRT-PCR-validated experimental data, achieving 98.4% accuracy and an area under the curve of 1. Our findings highlight this panel’s promising potential as a non-invasive approach not only for detecting HCC but also for distinguishing HCC from CHB patients.
Read full abstract