Objective: This study aims to identify and predict hub genes in the salivary transcriptome of oral cancer and healthy samples. Materials and methods: Salivary proteomic analysis was performed using samples from oral cancer patients and healthy controls, focusing on the parotid and submandibular glands. Gene set enrichment analysis (GSEA) was used to explore the enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG) and Gene Ontology (GO) pathways. Protein–protein interaction (PPI) networks were constructed using the STRING database and visualised in Cytoscape. Machine learning models, including naïve Bayes and neural networks, were applied to predict interactomic hub genes based on differentially expressed gene (DEG) data. Results: The machine learning models achieved an overall accuracy of 83% for the naïve Bayes classifier and 79% for the neural networks. Class-specific accuracies were 75% and 58%, respectively. Hub genes such as RACK1 and PON1 were identified as central interactomic players. The receiver operating characteristic curve demonstrated the model's capacity to differentiate between hub and non-hub genes, showcasing the potential for identifying critical biomarkers in oral cancer. Conclusions: The predictive accuracy of the naïve Bayes and neural network models underscores their potential in identifying key interactomic genes, which could improve treatment strategies and drug design.
Yadalam et al. (Tue,) studied this question.