Delineating novel diagnostic biomarkers and therapeutic targets for oral submucosal fibrosis: an integrative multi-omics and machine learning approach.

Background: Oral submucosal fibrosis (OSF) is a chronic and progressive disorder, caused by chewing areca nuts, affecting the oral cavity and oropharynx. OSF is characterized by severe symptoms like severe burning sensation, restricted mouth opening, etc. Given the multifactorial and poorly understood molecular basis of the disease, there is a need for novel biomarkers and therapeutic targets. Method: We downloaded 3 RNA-seq, two microarray, one epigenomic, and one single-cell RNA-seq datasets from the gene expression omnibus database. Differentially expressed genes (DEGs) were characterized using DESeq2. Several analyses, including gene enrichment, immune cell infiltration, protein-protein interaction, and more, were performed. Machine learning models were developed using all DEGs and top5 selected features with leave one out cross validation technique. Independent validations were performed using two microarray datasets with appropriate statistical measures. Epigenetic analysis revealed hyper- and hypomethylated genes based on delta-beta values, and an integrative analysis of the transcriptome and methylome was performed to obtain key biomarkers. Single-cell analysis was performed to identify key cell types showing higher DEG expression. Result: DESeq2 analysis identified 29 upregulated and 15 downregulated DEGs. Upregulated DEGs show enrichment for the inflammatory, metabolic, and signaling processes, whereas downregulated DEGs were largely associated with metabolic processes. Immune cell enrichment analysis using CIBERSORTx shows higher enrichment of "T cells," "mast cells," and "macrophages" in OSF patients. We validated our findings in two independent microarray datasets and observed a similar gene expression pattern of DEGs. Machine learning performed using top5 features where Random Forest model achieved AUROC of 0.99 and AUPRC of 0.99. Further, ROC analysis and AUC plot show that DEGs can discriminate OSF patients from the normal population with high AUROC. Integrative analysis of methylation and transcriptomic data identified 11 genes as potential diagnostic biomarkers and therapeutic targets. Finally, single-cell analysis elucidates the higher expression of DEGs in "keratinocyte", "epithelial cells" and "dendritic cells". Conclusion: Integrative analysis identified 11 gene signatures as potential early diagnostic biomarkers and therapeutic targets for the OSF. Furthermore, the study hints towards mechanistic insight into potential mechanism leading to oral cancer. All the codes and ML models are provided at our GitHub repository https://github.com/agrawalpiyush-srm/OSF.