Intelligent analysis of methylation data in Head and Neck Squamous Cell Carcinoma (HNSCC) interactomes

Abstract

Head and neck squamous cell carcinoma (HNSCC) is a heterogeneous disease which arises due to various genetic, epigenetic and environmental factors. DNA methylation is an epigenetic factor that is found to have a role in the development and progression of HNSCC through genetic and epigenetic silencing. Analysis of the methylation data can facilitate us to explore variations in several gene sites and can narrow down our search for curing HNSCC. The aim of this study was to explore and analyze the DNA Methylation data of HNSCC and make intelligent machine learning (ML) models that can predict the expression levels of a particular gene site based on various features. Difference between the gene expression levels of normal and tumor samples obtained from TCGA was calculated and then the genes were classified into hypo-methylated, hyper-methylated and non-methylated, respectively. Moreover, network analysis and functional enrichment analysis was performed to identify the protein-protein interaction (PPI) and involvement in the biological process followed by training logistic regression, support vector machine (SVM) and k-nearest neighbors (KNN) models for prediction. Logistic regression was found to have the highest accuracy of 65% among all the ML models. Furthermore, MYC, POLR2A, ALB, MTOR, H2AFX, SMARCA4, PAX6, GATA3 and MDM2 were identified as the hub genes in the HNSCC network. Whereas, hyper-methylated, hypo-methylated and non-methylated genes were found to be enriched in neuroactive ligand-receptor interaction, neurogenesis, ion transport channels, cell cycle and plasma membrane. In future, more data and features are required for validation and improving the accuracy of the ML models.