Prediction of key pathways in hepatocellular carcinoma (HCC): A machine learning approach using a sample pathway information matrix.

Hepatocellular carcinoma (HCC) represents the most prevalent form of primary liver cancer, accounting for 75 % of all cases. Individuals with metabolic dysfunctions are at risk of developing significant symptoms, including cirrhosis. To address this, we proposed a novel method to find the signalling pathway based on the patient's gene expression. The objective includes examining the predictive biomarkers associated with cirrhosis-related HCC. The study combines gene expression and pathway enrichment data to find the biologically important pathways in disease progression. The differential gene expression analysis showed 58 upregulated and 62 downregulating differentially expressed genes. These DEGs were utilized to construct a protein-protein interaction network, and then the clustered genes were determined. Subsequently, pathway enrichment analysis was performed for the clustered genes and the gene-pathway interaction matrix was developed. The sample-pathway information matrix (SPIM) was obtained by multiplying the gene-expression and gene-pathway matrix. The key pathways were predicted from the SPIM using random forest model and we achieved 94 % of accuracy. The arachidonic acid metabolism was the most important pathway and genes involved in this pathway includes CYP2C9, CYP2C8, CYP2B6. These genes are well known for promoting metabolic disorders in the liver. Hence, our novel method proves that it could distinguish the samples and extract important pathways that are involved in differentiating the diseased samples based on the gene expression. Therefore, integrating gene expression and their enriched biological pathways may effectively help in identifying the key signalling pathways.