Integrative Bioinformatics and Machine Learning Approaches for Biomarker and Pathway Discovery in Malignant Pleural Mesothelioma
Md. Lotifur Rohman, Suman BiswasABSTRACT
Background
Malignant pleural mesothelioma (MPM) is a highly aggressive thoracic cancer with limited treatment options and poor prognosis.
Aims
This study employed an integrative transcriptomic and computational approach to uncover potential biomarkers for MPM.
Methods
Gene expression data from the GSE51024 dataset, comprising 41 normal and 55 MPM tumor samples, were preprocessed using the robust multiarray average (RMA) method. Differentially expressed genes (DEGs) were identified using the Limma package and Wilcoxon rank tests. Weighted gene co‐expression network analysis (WGCNA) revealed significant gene modules using topological overlap and preservation statistics, followed by functional annotation through Gene Ontology (GO) and KEGG pathway analyses.
Results
Hub genes within conserved modules (MM > 0.8) were identified, including ZNF638, RAB18, SOS1, TTC17, RANBP2, FNDC3B, QSER1, CLDN15, PSTPIP1, SEPTIN6, ADAP2, and PVRIG. Survival analysis validated their prognostic relevance using public databases. Furthermore, high classification accuracy was achieved using various machine learning algorithms (SVM, PLDA, XGB, K‐NN, and logistic regression), confirming the discriminative potential of the hub genes. The results highlighted several physiologically significant hub genes and pathways associated with cell cycle regulation, immune modulation, and cancer progression.
Conclusion
Overall, this exploratory study demonstrated that potential biomarkers in MPM can be effectively identified using an integrative systems biology and machine learning framework. The identified hub genes may support future diagnostic, prognostic, and personalized treatment strategies for patients with MPM.