Identification and Validation of a Novel Immune Infiltration-Based Diagnostic Score for Early Detection of Hepatocellular Carcinoma by Machine-Learning Strategies.

Objective: To investigate the diagnostic gene biomarkers for hepatocellular carcinoma (HCC) and identify the immune cell infiltration characteristics in this pathology.

Methods: Five gene expression datasets were obtained through Gene Expression Omnibus (GEO) portal. After batch effect removal, differentially expressed genes (DEGs) were conducted between 209 HCC and 146 control tissues and functional correlation analyses were performed. Two machine learning algorithms were used to develop diagnostic signatures. The discriminatory ability of the gene signature was measured by AUC. The expression levels and diagnostic value of the identified biomarkers in HCC were further validated in three independent external cohorts. CIBERSORT algorithm was adopted to explore the immune infiltration of HCC. A correlation analysis was carried out between these diagnostic signatures and immune cells.

Results: A total of 375 DEGs were identified. GPC3, ACSM3, SPINK1, COL15A1, TP53I3, RRAGD, and CLDN10 were identified as the early diagnostic signatures of HCC and were all validated in external cohorts. The corresponding results of AUC presented excellent discriminatory ability of these feature genes. The immune cell infiltration analysis showed that multiple immune cells associated with these biomarkers may be involved in the development of HCC.

Conclusion: This study indicates that GPC3, ACSM3, SPINK1, COL15A1, TP53I3, RRAGD, and CLDN10 are potential biomarkers associated with immune infiltration in HCC. Combining these genes can be used for early detection of HCC and evaluating immune cell infiltration. Further studies are needed to explore their roles underlying the occurrence of HCC.